├── Sign_Language_Translation
    ├── __pycache__
    │   └── unicode.cpython-38.pyc
    ├── create_dataset_from_video.py
    ├── making_video.py
    ├── modules
    │   ├── __pycache__
    │   │   ├── holistic_module.cpython-38.pyc
    │   │   └── utils.cpython-38.pyc
    │   ├── holistic_module.py
    │   └── utils.py
    ├── readme.md
    ├── show_video.py
    ├── train_hand_gesture.ipynb
    ├── unicode.py
    ├── video_test_model_tflite.py
    └── webcam_test_model_tflite.py
├── dataset
    ├── output_video
    │   ├── ㄱ
    │   │   └── ㄱ_1.avi
    │   ├── ㄴ
    │   │   └── ㄴ_1.avi
    │   ├── ㄷ
    │   │   └── ㄷ_1.avi
    │   ├── ㄹ
    │   │   └── ㄹ_1.avi
    │   ├── ㅁ
    │   │   └── ㅁ_1.avi
    │   ├── ㅂ
    │   │   └── ㅂ_1.avi
    │   ├── ㅅ
    │   │   └── ㅅ_1.avi
    │   ├── ㅇ
    │   │   └── ㅇ_1.avi
    │   ├── ㅈ
    │   │   └── ㅈ_1.avi
    │   ├── ㅊ
    │   │   └── ㅊ_1.avi
    │   ├── ㅋ
    │   │   └── ㅋ_1.avi
    │   ├── ㅌ
    │   │   └── ㅌ_1.avi
    │   ├── ㅍ
    │   │   └── ㅍ_1.avi
    │   ├── ㅎ
    │   │   └── ㅎ_1.avi
    │   ├── ㅏ
    │   │   └── ㅏ_1.avi
    │   ├── ㅐ
    │   │   └── ㅐ_1.avi
    │   ├── ㅑ
    │   │   └── ㅑ_1.avi
    │   ├── ㅒ
    │   │   └── ㅒ_1.avi
    │   ├── ㅓ
    │   │   └── ㅓ_1.avi
    │   ├── ㅔ
    │   │   └── ㅔ_1.avi
    │   ├── ㅕ
    │   │   └── ㅕ_1.avi
    │   ├── ㅖ
    │   │   └── ㅖ_1.avi
    │   ├── ㅗ
    │   │   └── ㅗ_1.avi
    │   ├── ㅚ
    │   │   └── ㅚ_1.avi
    │   ├── ㅛ
    │   │   └── ㅛ_1.avi
    │   ├── ㅜ
    │   │   └── ㅜ_1.avi
    │   ├── ㅟ
    │   │   └── ㅟ_1.avi
    │   ├── ㅠ
    │   │   └── ㅠ_1.avi
    │   ├── ㅡ
    │   │   └── ㅡ_1.avi
    │   ├── ㅢ
    │   │   └── ㅢ_1.avi
    │   └── ㅣ
    │   │   └── ㅣ_1.avi
    ├── seq_ㄱ_1669720403.npy
    ├── seq_ㄱ_1669723415.npy
    ├── seq_ㄱ_1669724266.npy
    ├── seq_ㄴ_1669720403.npy
    ├── seq_ㄴ_1669723415.npy
    ├── seq_ㄴ_1669724266.npy
    ├── seq_ㄷ_1669720403.npy
    ├── seq_ㄷ_1669723415.npy
    ├── seq_ㄷ_1669724266.npy
    ├── seq_ㄹ_1669720403.npy
    ├── seq_ㄹ_1669723415.npy
    ├── seq_ㄹ_1669724266.npy
    ├── seq_ㅁ_1669720403.npy
    ├── seq_ㅁ_1669723415.npy
    ├── seq_ㅁ_1669724266.npy
    ├── seq_ㅂ_1669720403.npy
    ├── seq_ㅂ_1669723415.npy
    ├── seq_ㅂ_1669724266.npy
    ├── seq_ㅅ_1669720403.npy
    ├── seq_ㅅ_1669723415.npy
    ├── seq_ㅅ_1669724266.npy
    ├── seq_ㅇ_1669720403.npy
    ├── seq_ㅇ_1669723415.npy
    ├── seq_ㅇ_1669724266.npy
    ├── seq_ㅈ_1669720403.npy
    ├── seq_ㅈ_1669723415.npy
    ├── seq_ㅈ_1669724266.npy
    ├── seq_ㅊ_1669720403.npy
    ├── seq_ㅊ_1669723415.npy
    ├── seq_ㅊ_1669724266.npy
    ├── seq_ㅋ_1669720403.npy
    ├── seq_ㅋ_1669723415.npy
    ├── seq_ㅋ_1669724266.npy
    ├── seq_ㅌ_1669720403.npy
    ├── seq_ㅌ_1669723415.npy
    ├── seq_ㅌ_1669724266.npy
    ├── seq_ㅍ_1669720403.npy
    ├── seq_ㅍ_1669723415.npy
    ├── seq_ㅍ_1669724266.npy
    ├── seq_ㅎ_1669720403.npy
    ├── seq_ㅎ_1669723415.npy
    ├── seq_ㅎ_1669724266.npy
    ├── seq_ㅏ_1669720403.npy
    ├── seq_ㅏ_1669723415.npy
    ├── seq_ㅏ_1669724266.npy
    ├── seq_ㅐ_1669720403.npy
    ├── seq_ㅐ_1669723415.npy
    ├── seq_ㅐ_1669724266.npy
    ├── seq_ㅑ_1669720403.npy
    ├── seq_ㅑ_1669723415.npy
    ├── seq_ㅑ_1669724266.npy
    ├── seq_ㅒ_1669720403.npy
    ├── seq_ㅒ_1669723415.npy
    ├── seq_ㅒ_1669724266.npy
    ├── seq_ㅓ_1669720403.npy
    ├── seq_ㅓ_1669723415.npy
    ├── seq_ㅓ_1669724266.npy
    ├── seq_ㅔ_1669720403.npy
    ├── seq_ㅔ_1669723415.npy
    ├── seq_ㅔ_1669724266.npy
    ├── seq_ㅕ_1669720403.npy
    ├── seq_ㅕ_1669723415.npy
    ├── seq_ㅕ_1669724266.npy
    ├── seq_ㅖ_1669720403.npy
    ├── seq_ㅖ_1669723415.npy
    ├── seq_ㅖ_1669724266.npy
    ├── seq_ㅗ_1669720403.npy
    ├── seq_ㅗ_1669723415.npy
    ├── seq_ㅗ_1669724266.npy
    ├── seq_ㅚ_1669720403.npy
    ├── seq_ㅚ_1669723415.npy
    ├── seq_ㅚ_1669724266.npy
    ├── seq_ㅛ_1669720403.npy
    ├── seq_ㅛ_1669723415.npy
    ├── seq_ㅛ_1669724266.npy
    ├── seq_ㅜ_1669720403.npy
    ├── seq_ㅜ_1669723415.npy
    ├── seq_ㅜ_1669724266.npy
    ├── seq_ㅟ_1669720403.npy
    ├── seq_ㅟ_1669723415.npy
    ├── seq_ㅟ_1669724266.npy
    ├── seq_ㅠ_1669720403.npy
    ├── seq_ㅠ_1669723415.npy
    ├── seq_ㅠ_1669724266.npy
    ├── seq_ㅡ_1669720403.npy
    ├── seq_ㅡ_1669723415.npy
    ├── seq_ㅡ_1669724266.npy
    ├── seq_ㅢ_1669720403.npy
    ├── seq_ㅢ_1669723415.npy
    ├── seq_ㅢ_1669724266.npy
    ├── seq_ㅣ_1669720403.npy
    ├── seq_ㅣ_1669723415.npy
    └── seq_ㅣ_1669724266.npy
├── models
    ├── multi_hand_gesture_classifier.h5
    └── multi_hand_gesture_classifier.tflite
├── readme.md
├── version_requirements.txt
└── 실시간 수화 번역 인식 모듈 생성.pptx


/Sign_Language_Translation/__pycache__/unicode.cpython-38.pyc:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/Sign_Language_Translation/__pycache__/unicode.cpython-38.pyc


--------------------------------------------------------------------------------
/Sign_Language_Translation/create_dataset_from_video.py:
--------------------------------------------------------------------------------
  1 | import cv2
  2 | import sys, os
  3 | import mediapipe as mp
  4 | import numpy as np
  5 | import modules.holistic_module as hm
  6 | from modules.utils import createDirectory
  7 | import json
  8 | import time
  9 | 
 10 | from modules.utils import Vector_Normalization
 11 | 
 12 | createDirectory('dataset/output_video')
 13 | 
 14 | # 저장할 파일 이름
 15 | save_file_name = "train"
 16 | 
 17 | # 시퀀스의 길이(30 -> 10)
 18 | seq_length = 10
 19 | 
 20 | 
 21 | actions = ['ㄱ', 'ㄴ', 'ㄷ', 'ㄹ', 'ㅁ', 'ㅂ', 'ㅅ', 'ㅇ', 'ㅈ', 'ㅊ', 'ㅋ', 'ㅌ', 'ㅍ', 'ㅎ',
 22 |              'ㅏ', 'ㅑ', 'ㅓ', 'ㅕ', 'ㅗ', 'ㅛ', 'ㅜ', 'ㅠ', 'ㅡ', 'ㅣ',
 23 |              'ㅐ', 'ㅒ', 'ㅔ', 'ㅖ', 'ㅢ', 'ㅚ', 'ㅟ']
 24 | 
 25 | dataset = dict()
 26 | 
 27 | for i in range(len(actions)):
 28 |     dataset[i] = []
 29 | 
 30 | # MediaPipe holistic model
 31 | detector = hm.HolisticDetector(min_detection_confidence=0.3)
 32 | 
 33 | videoFolderPath = "dataset/output_video"
 34 | videoTestList = os.listdir(videoFolderPath)
 35 | 
 36 | testTargetList =[]
 37 | 
 38 | created_time = int(time.time())
 39 | 
 40 | for videoPath in videoTestList:
 41 |     actionVideoPath = f'{videoFolderPath}/{videoPath}'
 42 |     actionVideoList = os.listdir(actionVideoPath)
 43 |     for actionVideo in actionVideoList:
 44 |         fullVideoPath = f'{actionVideoPath}/{actionVideo}'
 45 |         testTargetList.append(fullVideoPath)
 46 | 
 47 | print("---------- Start Video List ----------")
 48 | testTargetList = sorted(testTargetList, key=lambda x:x[x.find("/", 9)+1], reverse=True)
 49 | print(testTargetList)
 50 | print("----------  End Video List  ----------\n")
 51 | 
 52 | for target in testTargetList:
 53 | 
 54 |     data = []
 55 |     first_index = target.find("/")
 56 |     second_index = target.find("/", first_index+1)
 57 |     third_index = target.find("/", second_index+1)
 58 |     idx = actions.index(target[target.find("/", second_index)+1:target.find("/", third_index)])
 59 | 
 60 |     print("Now Streaming :", target)
 61 |     cap = cv2.VideoCapture(target)
 62 | 
 63 |     # 열렸는지 확인
 64 |     if not cap.isOpened():
 65 |         print("Camera open failed!")
 66 |         sys.exit()
 67 | 
 68 |     # 웹캠의 속성 값을 받아오기
 69 |     # 정수 형태로 변환하기 위해 round
 70 |     w = round(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
 71 |     h = round(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
 72 |     fps = cap.get(cv2.CAP_PROP_FPS) # 카메라에 따라 값이 정상적, 비정상적
 73 |     print(w,h,'fps : ', fps)
 74 | 
 75 |     if fps != 0:
 76 |         delay = round(1000/fps)
 77 |     else:
 78 |         delay = round(1000/30)
 79 | 
 80 |     # 프레임을 받아와서 저장하기
 81 |     while True:
 82 |         ret, img = cap.read()
 83 | 
 84 |         if not ret:
 85 |             break
 86 | 
 87 |         img = detector.findHolistic(img, draw=True)
 88 |         # _, left_hand_lmList = detector.findLefthandLandmark(img)
 89 |         _, right_hand_lmList = detector.findRighthandLandmark(img)
 90 | 
 91 |         
 92 |         # if len(left_hand_lmList) != 0 and len(right_hand_lmList) != 0:
 93 |         # if left_hand_lmList is not None or right_hand_lmList is not None:
 94 |         if right_hand_lmList is not None:
 95 |             joint = np.zeros((42, 2)) # (21,2)
 96 |             
 97 |             # 왼손 랜드마크 리스트
 98 |             # for j, lm in enumerate(left_hand_lmList.landmark):
 99 |                 # joint[j] = [lm.x, lm.y]
100 |             
101 |             # 오른손 랜드마크 리스트
102 |             for j, lm in enumerate(right_hand_lmList.landmark):
103 |                 # joint[j+21] = [lm.x, lm.y]
104 |                 joint[j] = [lm.x, lm.y]
105 | 
106 |             # 벡터 정규화
107 |             vector, angle_label = Vector_Normalization(joint)
108 | 
109 |             # 정답 라벨링
110 |             angle_label = np.append(angle_label, idx)
111 | 
112 |             # 위치 종속성을 가지는 데이터 저장
113 |             # d = np.concatenate([joint.flatten(), angle_label])
114 | 
115 |             # 벡터 정규화를 활용한 위치 종속성 제거
116 |             d = np.concatenate([vector.flatten(), angle_label.flatten()])
117 |             
118 |             data.append(d)
119 | 
120 |         
121 | 
122 |         # draw box
123 |         # cv2.rectangle(img, (0,0), (w, 30), (245, 117, 16), -1)
124 | 
125 |         # draw text target name
126 |         # cv2.putText(img, target, (15,20), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 0), 1, cv2.LINE_AA)
127 | 
128 | 
129 |         # cv2.imshow('img', img)
130 |         cv2.waitKey(delay)
131 | 
132 |         # esc를 누르면 강제 종료
133 |         if cv2.waitKey(delay) == 27: 
134 |             break
135 | 
136 |     print("\n---------- Finish Video Streaming ----------")
137 | 
138 |     data = np.array(data)
139 | 
140 |     # Create sequence data
141 |     print('len(data)-seq_lenth:', len(data) - seq_length)
142 |     for seq in range(len(data) - seq_length):
143 |         dataset[idx].append(data[seq:seq + seq_length])    
144 | 
145 | '''
146 | for i in range(len(actions)):
147 |     save_data = np.array(dataset[i])
148 |     np.save(os.path.join('dataset', f'seq_{actions[i]}_{created_time}'), save_data)
149 | 
150 | 
151 | print("\n---------- Finish Save Dataset ----------")
152 | '''
153 | 
154 | 
155 | 
156 | 


--------------------------------------------------------------------------------
/Sign_Language_Translation/making_video.py:
--------------------------------------------------------------------------------
  1 | import cv2
  2 | import sys, os
  3 | import time
  4 | import mediapipe as mp
  5 | from modules.utils import createDirectory
  6 | import numpy as np
  7 | from PIL import ImageFont, ImageDraw, Image
  8 | 
  9 | fontpath = "fonts/HMKMMAG.TTF"
 10 | font = ImageFont.truetype(fontpath, 40)
 11 | 
 12 | createDirectory('dataset')
 13 | 
 14 | # 손가락 통증 이슈로 나누어 찍기
 15 | # actions = ['ㄱ', 'ㄴ', 'ㄷ', 'ㄹ', 'ㅁ', 'ㅂ', 'ㅅ', 'ㅇ', 'ㅈ', 'ㅊ', 'ㅋ', 'ㅌ', 'ㅍ', 'ㅎ']
 16 | # actions = ['ㅏ', 'ㅑ', 'ㅓ', 'ㅕ', 'ㅗ', 'ㅛ', 'ㅜ', 'ㅠ', 'ㅡ', 'ㅣ']
 17 | # actions = ['ㅗ', 'ㅛ', 'ㅜ']
 18 | actions = ['ㅐ', 'ㅒ', 'ㅔ', 'ㅖ', 'ㅢ', 'ㅚ', 'ㅟ']
 19 | # seq_length = 10
 20 | secs_for_action = 30
 21 | 
 22 | # MediaPipe hands model
 23 | mp_hands = mp.solutions.hands
 24 | mp_drawing = mp.solutions.drawing_utils
 25 | hands = mp_hands.Hands(
 26 |     max_num_hands=1,
 27 |     min_detection_confidence=0.3,
 28 |     min_tracking_confidence=0.5)
 29 | 
 30 | cap = cv2.VideoCapture(0)
 31 | 
 32 | created_time = int(time.time())
 33 | 
 34 | 
 35 | # 열렸는지 확인
 36 | if not cap.isOpened():
 37 |     print("Camera open failed!")
 38 |     sys.exit()
 39 | 
 40 | # 웹캠의 속성 값을 받아오기
 41 | # 정수 형태로 변환하기 위해 round
 42 | w = round(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
 43 | h = round(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
 44 | fps = cap.get(cv2.CAP_PROP_FPS) # 카메라에 따라 값이 정상적, 비정상적
 45 | 
 46 | if fps != 0:
 47 |     delay = round(1000/fps)
 48 | else:
 49 |     delay = round(1000/30)
 50 | 
 51 | fourcc = cv2.VideoWriter_fourcc(*'DIVX')
 52 | 
 53 | 
 54 | # 프레임을 받아와서 저장하기
 55 | while cap.isOpened():
 56 |     for idx, action in enumerate(actions):
 57 |         
 58 |         os.makedirs(f'dataset/output_video/{action}', exist_ok=True)
 59 | 
 60 |         videoFolderPath = f'dataset/output_video/{action}'
 61 |         videoList = sorted(os.listdir(videoFolderPath), key=lambda x:int(x[x.find("_")+1:x.find(".")]))
 62 |       
 63 |         if len(videoList) == 0:
 64 |             take = 1
 65 |         else:
 66 |             f = videoList[-1].find("_")
 67 |             e = videoList[-1].find(".")
 68 |             take = int(videoList[-1][f+1:e]) + 1
 69 | 
 70 |         saved_video_path = f'dataset/output_video/{action}/{action}_{take}.avi'
 71 | 
 72 |         out = cv2.VideoWriter(saved_video_path, fourcc, fps, (w, h))
 73 | 
 74 |         ret, img = cap.read()
 75 |         if not ret:
 76 |             break
 77 |          
 78 |         # 한글 폰트 출력    
 79 |         img_pil = Image.fromarray(img)
 80 |         draw = ImageDraw.Draw(img_pil)
 81 |         draw.text((10, 30), f'{action.upper()} 입력 대기중..', font=font, fill=(255, 255, 255))
 82 |         img = np.array(img_pil)
 83 | 
 84 |         #cv2.putText(img, f'Waiting for collecting {action.upper()} action...', org=(10, 30), fontFace=cv2.FONT_HERSHEY_SIMPLEX, fontScale=1, color=(255, 255, 255), thickness=2)
 85 |         cv2.imshow('img', img)
 86 |         cv2.waitKey(4000)
 87 | 
 88 |         start_time = time.time()
 89 | 
 90 |         while time.time() - start_time < secs_for_action:
 91 |             ret, img = cap.read()
 92 |             if not ret:
 93 |                 break
 94 |             
 95 |             # 비디오 녹화
 96 |             out.write(img)
 97 | 
 98 |             img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
 99 |             result = hands.process(img)
100 |             img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
101 | 
102 |             if result.multi_hand_landmarks is not None:
103 |                 for res in result.multi_hand_landmarks:
104 |                     mp_drawing.draw_landmarks(img, res, mp_hands.HAND_CONNECTIONS)
105 | 
106 |             cv2.imshow('img', img)
107 | 
108 |             # esc를 누르면 강제 종료
109 |             if cv2.waitKey(delay) == 27: 
110 |                 break
111 | 
112 |     cap.release()
113 |     out.release()
114 |     cv2.destroyAllWindows()


--------------------------------------------------------------------------------
/Sign_Language_Translation/modules/__pycache__/holistic_module.cpython-38.pyc:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/Sign_Language_Translation/modules/__pycache__/holistic_module.cpython-38.pyc


--------------------------------------------------------------------------------
/Sign_Language_Translation/modules/__pycache__/utils.cpython-38.pyc:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/Sign_Language_Translation/modules/__pycache__/utils.cpython-38.pyc


--------------------------------------------------------------------------------
/Sign_Language_Translation/modules/holistic_module.py:
--------------------------------------------------------------------------------
  1 | import cv2
  2 | import mediapipe as mp
  3 | import time
  4 | import math
  5 | 
  6 | class HolisticDetector():
  7 |     def __init__(self,
  8 |                static_image_mode=False,
  9 |                model_complexity=1,
 10 |                smooth_landmarks=True,
 11 |                enable_segmentation=False,
 12 |                smooth_segmentation=True,
 13 |                refine_face_landmarks=False,
 14 |                min_detection_confidence=0.5,
 15 |                min_tracking_confidence=0.5):
 16 |         self.static_image_mode = static_image_mode
 17 |         self.model_complexity = model_complexity
 18 |         self.smooth_landmarks = smooth_landmarks
 19 |         self.enable_segmentation = enable_segmentation
 20 |         self.smooth_segmentation = smooth_segmentation
 21 |         self.refine_face_landmarks = refine_face_landmarks
 22 |         self.min_detection_confidence = min_detection_confidence
 23 |         self.min_tracking_confidence = min_tracking_confidence
 24 | 
 25 |         self.mpHolistic = mp.solutions.holistic
 26 |         self.mpPose = mp.solutions.pose
 27 |         self.mpFace = mp.solutions.face_mesh
 28 |         self.holistics = self.mpHolistic.Holistic(self.static_image_mode,
 29 |                                                   self.model_complexity,
 30 |                                                   self.smooth_landmarks,
 31 |                                                   self.enable_segmentation,
 32 |                                                   self.smooth_segmentation,
 33 |                                                   self.refine_face_landmarks,
 34 |                                                   self.min_detection_confidence,
 35 |                                                   self.min_tracking_confidence)
 36 |         self.mpDraw = mp.solutions.drawing_utils
 37 | 
 38 |         self.tipIds = [4, 8, 12, 16, 20]
 39 | 
 40 |     def findHolistic(self, img, draw = True):
 41 |         imgRGB = cv2.cvtColor(img, cv2.COLOR_BGR2RGB )
 42 |         self.results = self.holistics.process(imgRGB)
 43 | 
 44 |         if self.results.pose_landmarks:
 45 |             
 46 |             if draw:
 47 |                  # Draw pose, left and right hands, and face landmarks on the image.
 48 |                 annotated_image = img.copy()
 49 | 
 50 |                 # self.mpDraw.draw_landmarks(
 51 |                 #     annotated_image, self.results.face_landmarks, self.mpHolistic.FACE_CONNECTIONS)
 52 |                 self.mpDraw.draw_landmarks(
 53 |                     annotated_image, self.results.left_hand_landmarks, self.mpHolistic.HAND_CONNECTIONS)
 54 |                 self.mpDraw.draw_landmarks(
 55 |                     annotated_image, self.results.right_hand_landmarks, self.mpHolistic.HAND_CONNECTIONS)
 56 |                 # self.mpDraw.draw_landmarks(
 57 |                 #     annotated_image, self.results.pose_landmarks, self.mpHolistic.POSE_CONNECTIONS)
 58 | 
 59 |                 # Plot pose world landmarks.
 60 |                 # self.mpDraw.plot_landmarks(
 61 |                 #     self.results.pose_world_landmarks, self.mpHolistic.POSE_CONNECTIONS)
 62 |                 return annotated_image
 63 |             
 64 |         return img
 65 | 
 66 |     def findPoseLandmark(self, img, draw=True):
 67 |         xList = []
 68 |         yList = []
 69 | 
 70 |         self.pose_lmList = []
 71 |         if self.results.pose_landmarks:
 72 |             myHolistic = self.results.pose_landmarks
 73 |             # print(myHolistic.landmark)
 74 |             # print(type(myHolistic.landmark))
 75 |             for id, lm in enumerate(myHolistic.landmark):
 76 |                 # print(id,lm)
 77 |                 h, w, c = img.shape
 78 |                 cx, cy, cz = int(lm.x*w), int(lm.y*h), int(lm.z*(w+h)/2)
 79 |                 # print(id, cx, cy)
 80 |                 # print(cz)
 81 |                 xList.append(cx)
 82 |                 yList.append(cy)
 83 |                 self.pose_lmList.append([id, cx, cy, cz])
 84 | 
 85 |         return self.pose_lmList
 86 | 
 87 |     def findFaceLandmark(self, img, draw=True):
 88 |         xList = []
 89 |         yList = []
 90 | 
 91 |         self.face_lmList = []
 92 |         if self.results.face_landmarks:
 93 |             myHolistic = self.results.face_landmarks
 94 |             # print(type(myHolistic.landmark))
 95 |             for id, lm in enumerate(myHolistic.landmark):
 96 |                 # print(id,lm)
 97 |                 h, w, c = img.shape
 98 |                 cx, cy, cz = int(lm.x*w), int(lm.y*h), int(lm.z*(w+h)/2)
 99 |                 # print(id, cx, cy)
100 |                 xList.append(cx)
101 |                 yList.append(cy)
102 |                 self.face_lmList.append([id, cx, cy, cz])
103 | 
104 |         return self.face_lmList
105 | 
106 |     def findLefthandLandmark(self, img, draw=True):
107 |         xList = []
108 |         yList = []
109 | 
110 |         self.left_hand_lmList = []
111 |         if self.results.left_hand_landmarks:
112 |             myHolistic = self.results.left_hand_landmarks
113 |             
114 |             for id, lm in enumerate(myHolistic.landmark):
115 |                 h, w, c = img.shape
116 |                 cx, cy, cz = int(lm.x*w), int(lm.y*h), int(lm.z*(w+h)/2)
117 |                 xList.append(cx)
118 |                 yList.append(cy)
119 |                 self.left_hand_lmList.append([id, cx, cy, cz])
120 | 
121 |         return self.left_hand_lmList, self.results.left_hand_landmarks
122 | 
123 |     def findRighthandLandmark(self, img, draw=True):
124 |         xList = []
125 |         yList = []
126 | 
127 |         self.right_hand_lmList = []
128 |         if self.results.right_hand_landmarks:
129 |             myHolistic = self.results.right_hand_landmarks
130 |             for id, lm in enumerate(myHolistic.landmark):
131 |                 h, w, c = img.shape
132 |                 cx, cy, cz = int(lm.x*w), int(lm.y*h), int(lm.z*(w+h)/2)
133 |                 xList.append(cx)
134 |                 yList.append(cy)
135 |                 self.right_hand_lmList.append([id, cx, cy, cz])
136 | 
137 |         return self.right_hand_lmList, self.results.right_hand_landmarks
138 | 
139 |     def left_hand_fingersUp(self, axis=False):
140 |         fingers = []
141 |         
142 |         if axis == False:
143 |             # Thumb
144 |             if self.left_hand_lmList[self.tipIds[0]][1] < self.left_hand_lmList[self.tipIds[4]][1]:
145 |                 if self.left_hand_lmList[self.tipIds[0]][1] < self.left_hand_lmList[self.tipIds[0] - 2][1]:
146 |                     fingers.append(1)
147 |                 else:
148 |                     fingers.append(0)
149 |             elif self.left_hand_lmList[self.tipIds[0]][1] > self.left_hand_lmList[self.tipIds[4]][1]:
150 |                 if self.left_hand_lmList[self.tipIds[0]][1] > self.left_hand_lmList[self.tipIds[0] - 2][1]:
151 |                     fingers.append(1)
152 |                 else:
153 |                     fingers.append(0)
154 | 
155 |             # Fingers except Thumb
156 |             for id in range(1, 5):
157 |                 if self.left_hand_lmList[self.tipIds[id]][2] < self.left_hand_lmList[self.tipIds[id]-2][2]:
158 |                     fingers.append(1)
159 |                 else:
160 |                     fingers.append(0)
161 | 
162 |         # axis = True( to detect LIKE gesture )
163 |         else:
164 |             # Thumb
165 |             if self.left_hand_lmList[self.tipIds[0]][2] < self.left_hand_lmList[self.tipIds[0] - 2][2]:
166 |                 fingers.append(1)
167 |             else:
168 |                 fingers.append(0)
169 | 
170 |             # Fingers except Thumb
171 |             if self.left_hand_lmList[self.tipIds[0]][1] < self.left_hand_lmList[self.tipIds[4]][1]:
172 |                 for id in range(1, 5):
173 |                     if self.left_hand_lmList[self.tipIds[id]][1] > self.left_hand_lmList[self.tipIds[id]-2][1]:
174 |                         fingers.append(1)
175 |                     else:
176 |                         fingers.append(0)
177 |             else:
178 |                 for id in range(1, 5):
179 |                     if self.left_hand_lmList[self.tipIds[id]][1] < self.left_hand_lmList[self.tipIds[id]-2][1]:
180 |                         fingers.append(1)
181 |                     else:
182 |                         fingers.append(0)
183 | 
184 | 
185 |         return fingers
186 | 
187 |     def right_hand_fingersUp(self, axis=False):
188 |         fingers = []
189 | 
190 |         if axis == False:
191 |             # Thumb
192 |             if self.right_hand_lmList[self.tipIds[0]][1] > self.right_hand_lmList[self.tipIds[4]][1]:
193 |                 if self.right_hand_lmList[self.tipIds[0]][1] > self.right_hand_lmList[self.tipIds[0] - 2][1]:
194 |                     fingers.append(1)
195 |                 else:
196 |                     fingers.append(0)
197 |             if self.right_hand_lmList[self.tipIds[0]][1] < self.right_hand_lmList[self.tipIds[4]][1]:
198 |                 if self.right_hand_lmList[self.tipIds[0]][1] < self.right_hand_lmList[self.tipIds[0] - 2][1]:
199 |                     fingers.append(1)
200 |                 else:
201 |                     fingers.append(0)
202 | 
203 |             # Fingers except Thumb
204 |             for id in range(1, 5):
205 |                 if self.right_hand_lmList[self.tipIds[id]][2] < self.right_hand_lmList[self.tipIds[id]-2][2]:
206 |                     fingers.append(1)
207 |                 else:
208 |                     fingers.append(0)
209 | 
210 |         # axis = True( to detect LIKE gesture )
211 |         else:
212 |             # Thumb
213 |             if self.right_hand_lmList[self.tipIds[0]][2] < self.right_hand_lmList[self.tipIds[0] - 2][2]:
214 |                 fingers.append(1)
215 |             else:
216 |                 fingers.append(0)
217 | 
218 |             # Fingers except Thumb
219 |             if self.right_hand_lmList[self.tipIds[0]][1] < self.right_hand_lmList[self.tipIds[4]][1]:
220 |                 for id in range(1, 5):
221 |                     if self.right_hand_lmList[self.tipIds[id]][1] > self.right_hand_lmList[self.tipIds[id]-2][1]:
222 |                         fingers.append(1)
223 |                     else:
224 |                         fingers.append(0)
225 |             else:
226 |                 for id in range(1, 5):
227 |                     if self.right_hand_lmList[self.tipIds[id]][1] < self.right_hand_lmList[self.tipIds[id]-2][1]:
228 |                         fingers.append(1)
229 |                     else:
230 |                         fingers.append(0)
231 | 
232 |         return fingers
233 | 
234 |     def findCenter(self, p1, p2):
235 |         x1, y1 = self.pose_lmList[p1][1:3]
236 |         x2, y2 = self.pose_lmList[p2][1:3]
237 |         cx, cy = (x1 + x2) // 2, (y1 + y2) // 2
238 | 
239 |         return cx, cy
240 | 
241 |     def findDistance(self, p1, p2, img, draw=True, r=15, t=3):
242 |         x1, y1 = self.face_lmList[p1][1:3]
243 |         x2, y2 = p2[0],p2[1]
244 | 
245 |         if draw:
246 |             cv2.line(img, (x1, y1), (x2, y2), (255,0,255), t)
247 |             cv2.circle(img, (x1, y1), r, (255,0,255), cv2.FILLED)
248 |             cv2.circle(img, (x2, y2), r, (255,0,255), cv2.FILLED)
249 |         length = math.hypot(x2-x1, y2-y1)
250 | 
251 |         return length, img
252 | 
253 |     def findDepth(self, p1, p2):
254 |         depth = abs((self.pose_lmList[p1][3] + self.pose_lmList[p2][3]) / 2)
255 |         return depth
256 | 
257 |     def findEyeBlink(self, p1, p2, img, draw=True, r=15, t=3):
258 |         x1, y1 = self.face_lmList[p1][1:3]
259 |         x2, y2 = self.face_lmList[p2][1:3]
260 | 
261 |         if draw:
262 |             cv2.line(img, (x1, y1), (x2, y2), (255,0,255), t)
263 |             cv2.circle(img, (x1, y1), r, (255,0,255), cv2.FILLED)
264 |             cv2.circle(img, (x2, y2), r, (255,0,255), cv2.FILLED)
265 |         length = math.hypot(x2-x1, y2-y1)
266 | 
267 |         return length, img
268 | 
269 |     def findEyeDepth(self, p1, p2):
270 |         depth = abs((self.face_lmList[p1][3] + self.face_lmList[p2][3]) / 2)
271 |         return depth
272 | 
273 |     def drawLine(self, p1, p2, img, t=3):
274 |         x1, y1 = self.face_lmList[p1][1:3]
275 |         x2, y2 = self.face_lmList[p2][1:3]
276 | 
277 |         cv2.line(img, (x1, y1), (x2, y2), (255,255,255), t)
278 | 
279 |     def findLength_lh_rh(self, p1, p2):
280 |         x1, y1 = self.left_hand_lmList[p2][1:3]
281 |         x2, y2 = self.right_hand_lmList[p1][1:3]
282 | 
283 |         length = math.hypot(abs(x2-x1), abs(y2-y1))
284 |         return length
285 | 
286 |     def findLength_lh_lh(self, p1, p2):
287 |         x1, y1 = self.left_hand_lmList[p2][1:3]
288 |         x2, y2 = self.left_hand_lmList[p1][1:3]
289 | 
290 |         length = math.hypot(abs(x2-x1), abs(y2-y1))
291 |         return length
292 | 
293 |     def findLength_rh_rh(self, p1, p2):
294 |         x1, y1 = self.right_hand_lmList[p2][1:3]
295 |         x2, y2 = self.right_hand_lmList[p1][1:3]
296 | 
297 |         length = math.hypot(abs(x2-x1), abs(y2-y1))
298 |         return length
299 | 
300 |     def findLength_pose(self, p1, p2):
301 |         x1, y1 = self.pose_lmList[p2][1:3]
302 |         x2, y2 = self.pose_lmList[p1][1:3]
303 | 
304 |         length = math.hypot(abs(x2-x1), abs(y2-y1))
305 |         return length
306 | 
307 |     def findAngle(self, img, p1, p2, p3, draw=True):
308 |         # 랜드마크 좌표 얻기
309 |         # , x1, y1 = self.lmList[p1]
310 |         x1, y1 = self.pose_lmList[p1][1:3]
311 |         x2, y2 = self.pose_lmList[p2][1:3]
312 |         x3, y3 = self.pose_lmList[p3][1:3]
313 | 
314 |         # 각도 계산
315 |         radian = math.atan2(y3-y2,x3-x2)-math.atan2(y1-y2,x1-x2)
316 |         angle = math.degrees(radian)
317 | 
318 |         if angle < 0:
319 |             angle += 360
320 | 
321 |         #print(angle)
322 |         # 점, 선 그리기
323 |         if draw:
324 |             cv2.line(img, (x1,y1), (x2,y2), (255,255,255), 3)
325 |             cv2.line(img, (x2,y2), (x3,y3), (255,255,255), 3)            
326 |             cv2.circle(img, (x1,y1), 10, (0,0,255), cv2.FILLED)
327 |             cv2.circle(img, (x1,y1), 15, (0,0,255), 2)
328 |             cv2.circle(img, (x2,y2), 10, (0,0,255), cv2.FILLED)
329 |             cv2.circle(img, (x2,y2), 15, (0,0,255), 2)
330 |             cv2.circle(img, (x3,y3), 10, (0,0,255), cv2.FILLED)
331 |             cv2.circle(img, (x3,y3), 15, (0,0,255), 2)
332 |             cv2.putText(img, str(int(angle)), (x2-50,y2+50), cv2.FONT_HERSHEY_PLAIN, 2, (0,0,255), 2) 
333 |             
334 |         return angle
335 | 
336 |     def findHandAngle(self, img, p1, p2, p3, draw=True):
337 |         # 랜드마크 좌표 얻기
338 |         # , x1, y1 = self.lmList[p1]
339 |         x1, y1 = self.right_hand_lmList[p1][1:3]
340 |         x2, y2 = self.right_hand_lmList[p2][1:3]
341 |         x3, y3 = self.right_hand_lmList[p3][1:3]
342 | 
343 |         # 각도 계산
344 |         radian = math.atan2(y3-y2,x3-x2)-math.atan2(y1-y2,x1-x2)
345 |         angle = math.degrees(radian)
346 | 
347 |         if angle < 0:
348 |             angle += 360
349 | 
350 |         #print(angle)
351 |         # 점, 선 그리기
352 |         if draw:
353 |             cv2.line(img, (x1,y1), (x2,y2), (255,255,255), 3)
354 |             cv2.line(img, (x2,y2), (x3,y3), (255,255,255), 3)            
355 |             cv2.circle(img, (x1,y1), 10, (0,0,255), cv2.FILLED)
356 |             cv2.circle(img, (x1,y1), 15, (0,0,255), 2)
357 |             cv2.circle(img, (x2,y2), 10, (0,0,255), cv2.FILLED)
358 |             cv2.circle(img, (x2,y2), 15, (0,0,255), 2)
359 |             cv2.circle(img, (x3,y3), 10, (0,0,255), cv2.FILLED)
360 |             cv2.circle(img, (x3,y3), 15, (0,0,255), 2)
361 |             cv2.putText(img, str(int(angle)), (x2-50,y2+50), cv2.FONT_HERSHEY_PLAIN, 2, (0,0,255), 2) 
362 |             
363 |         return angle


--------------------------------------------------------------------------------
/Sign_Language_Translation/modules/utils.py:
--------------------------------------------------------------------------------
 1 | import os
 2 | import cv2
 3 | import numpy as np
 4 | 
 5 | def createDirectory(directory):
 6 |     try:
 7 |         if not os.path.exists(directory):
 8 |             os.makedirs(directory)
 9 |     except OSError:
10 |         print("Error: Failed to create the directory.")
11 | 
12 | # vector normalization
13 | def Vector_Normalization(joint):
14 |     # Compute angles between joints
15 |     v1 = joint[[0,1,2,3,0,5,6,7,0,9,10,11,0,13,14,15,0,17,18,19], :2] # Parent joint
16 |     v2 = joint[[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20], :2] # Child joint
17 |     v = v2 - v1 
18 |     # Normalize v
19 |     v = v / np.linalg.norm(v, axis=1)[:, np.newaxis]
20 | 
21 |     # Get angle using arcos of dot product
22 |     angle = np.arccos(np.einsum('nt,nt->n',
23 |         v[[0,1,2,4,5,6,8,9,10,12,13,14,16,17,18],:], 
24 |         v[[1,2,3,5,6,7,9,10,11,13,14,15,17,18,19],:])) 
25 | 
26 |     angle = np.degrees(angle) # Convert radian to degree
27 | 
28 |     angle_label = np.array([angle], dtype=np.float32)
29 | 
30 |     return v, angle_label


--------------------------------------------------------------------------------
/Sign_Language_Translation/readme.md:
--------------------------------------------------------------------------------
 1 | ### User Guide [예제 동작 순서]
 2 | 
 3 | 1. making_video.py
 4 | - 원하는 양 손 제스쳐 동영상을 촬영합니다.
 5 | - 동영상 촬영은 30초를 기준으로 하고 중단을 원하면, "ESC" 키를 눌러 종료합니다.
 6 | - 촬영을 여러번 진행할 수 있습니다.
 7 | - [주의] 촬영 중 인식되는 랜드마크는 동영상에 저장되지 않습니다.
 8 | 
 9 | 2. show_video.py (생략 가능)
10 | - 촬영한 데이터를 확인합니다.
11 | - video_path를 지정하면, 경로의 하위 모든 비디오 파일을 재생합니다.
12 | 
13 | 3. create_dataset_from_video.py
14 | - 촬영한 데이터를 활용하여 양손 관절 및 각도를 시퀀스 데이터로 변환하여 npy 파일로 저장합니다.
15 | 
16 | 4. train_hand_gesture.ipynb
17 | - npy file load하여 모델을 생성합니다.
18 | - keras model, tflite model 두 종류의 모델을 생성합니다. (tflite model만 활용)
19 | 
20 | 5. video_test_model : videoFolderPath를 지정하여 비디오를 활용하여 테스트합니다.
21 | - tflite.py : tflite model을 테스트합니다.
22 | 
23 | 6. webcam_test_model : 웹캠을 활용하여 테스트합니다.
24 | - tflite.py : tflite model을 테스트합니다.
25 | 
26 | c.f) unicode.py는 자음모음 결합을 위해 사용하려 했으나 사용 x
27 | 
28 | 
29 | <br>
30 | 
31 | ### 데이터 정규화 방법 선택
32 | - 필요에 따라 데이터 정규화를 선택 적용합니다. (적용 or 적용 X)
33 | - 2가지 코드 중 1개만 선택합니다.
34 | 
35 | ```python
36 | # 위치 종속성을 가지는 데이터 저장
37 | d = np.concatenate([joint.flatten(), angle_label])
38 | 
39 | # 정규화 벡터를 활용한 위치 종속성 제거
40 | d = np.concatenate([vector.flatten(), angle_label.flatten()])
41 | 
42 | ```
43 | (Examples) 벡터 정규화를 적용하는 코드
44 | ```python
45 | # 위치 종속성을 가지는 데이터 저장
46 | # d = np.concatenate([joint.flatten(), angle_label])
47 | 
48 | # 정규화 벡터를 활용한 위치 종속성 제거
49 | d = np.concatenate([vector.flatten(), angle_label.flatten()])
50 | 
51 | ```


--------------------------------------------------------------------------------
/Sign_Language_Translation/show_video.py:
--------------------------------------------------------------------------------
 1 | import cv2
 2 | import sys, os
 3 | import numpy as np
 4 | from PIL import ImageFont, ImageDraw, Image
 5 | 
 6 | fontpath = "fonts/HMKMMAG.TTF"
 7 | font = ImageFont.truetype(fontpath, 40)
 8 | 
 9 | videoFolderPath = "dataset/output_video"
10 | videoTestList = os.listdir(videoFolderPath)
11 | 
12 | testTargetList =[]
13 | 
14 | for videoPath in videoTestList:
15 |     actionVideoPath = f'{videoFolderPath}/{videoPath}'
16 |     actionVideoList = os.listdir(actionVideoPath)
17 |     for actionVideo in actionVideoList:
18 |         fullVideoPath = f'{actionVideoPath}/{actionVideo}'
19 |         testTargetList.append(fullVideoPath)
20 | 
21 | print("---------- Start Video List ----------")
22 | testTargetList = sorted(testTargetList, key=lambda x:x[x.find("/", 9)+1], reverse=True)
23 | print(testTargetList)
24 | print("----------  End Video List  ----------\n")
25 | 
26 | for target in testTargetList:
27 |     print("Now Streaming :", target)
28 |     cap = cv2.VideoCapture(target)
29 | 
30 |     # 열렸는지 확인
31 |     if not cap.isOpened():
32 |         print("Camera open failed!")
33 |         sys.exit()
34 | 
35 |     # 웹캠의 속성 값을 받아오기
36 |     # 정수 형태로 변환하기 위해 round
37 |     w = round(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
38 |     h = round(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
39 |     fps = cap.get(cv2.CAP_PROP_FPS) # 카메라에 따라 값이 정상적, 비정상적
40 | 
41 |     if fps != 0:
42 |         delay = round(1000/fps)
43 |     else:
44 |         delay = round(1000/30)
45 | 
46 |     # 프레임을 받아와서 저장하기
47 |     while True:
48 |         ret, img = cap.read()
49 | 
50 |         if not ret:
51 |             break
52 | 
53 |         # draw box
54 |         cv2.rectangle(img, (0,0), (w, 70), (245, 117, 16), -1)
55 | 
56 |         # draw text target name
57 |         #cv2.putText(img, target, (15,20), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 0), 1, cv2.LINE_AA)
58 | 
59 |         # 한글 폰트 출력    
60 |         img_pil = Image.fromarray(img)
61 |         draw = ImageDraw.Draw(img_pil)
62 |         draw.text((15,20), target[23:], font=font, fill=(0, 0, 0))
63 |         img = np.array(img_pil)
64 | 
65 | 
66 |         cv2.imshow('img', img)
67 |         cv2.waitKey(delay)
68 | 
69 |         # esc를 누르면 강제 종료
70 |         if cv2.waitKey(delay) == 27: 
71 |             break
72 | 
73 | 
74 |     cap.release()
75 |     cv2.destroyAllWindows()
76 | print("\n---------- Finish Video Streaming ----------")


--------------------------------------------------------------------------------
/Sign_Language_Translation/unicode.py:
--------------------------------------------------------------------------------
  1 | __all__ = ["split_syllable_char", "split_syllables",
  2 |            "join_jamos", "join_jamos_char",
  3 |            "CHAR_INITIALS", "CHAR_MEDIALS", "CHAR_FINALS"]
  4 | 
  5 | # 자음모음 결합용
  6 | import itertools
  7 | 
  8 | INITIAL = 0x001
  9 | MEDIAL = 0x010
 10 | FINAL = 0x100
 11 | CHAR_LISTS = {
 12 |     INITIAL: list(map(chr, [
 13 |         0x3131, 0x3132, 0x3134, 0x3137, 0x3138, 0x3139,
 14 |         0x3141, 0x3142, 0x3143, 0x3145, 0x3146, 0x3147,
 15 |         0x3148, 0x3149, 0x314a, 0x314b, 0x314c, 0x314d,
 16 |         0x314e
 17 |     ])),
 18 |     MEDIAL: list(map(chr, [
 19 |         0x314f, 0x3150, 0x3151, 0x3152, 0x3153, 0x3154,
 20 |         0x3155, 0x3156, 0x3157, 0x3158, 0x3159, 0x315a,
 21 |         0x315b, 0x315c, 0x315d, 0x315e, 0x315f, 0x3160,
 22 |         0x3161, 0x3162, 0x3163
 23 |     ])),
 24 |     FINAL: list(map(chr, [
 25 |         0x3131, 0x3132, 0x3133, 0x3134, 0x3135, 0x3136,
 26 |         0x3137, 0x3139, 0x313a, 0x313b, 0x313c, 0x313d,
 27 |         0x313e, 0x313f, 0x3140, 0x3141, 0x3142, 0x3144,
 28 |         0x3145, 0x3146, 0x3147, 0x3148, 0x314a, 0x314b,
 29 |         0x314c, 0x314d, 0x314e
 30 |     ]))
 31 | }
 32 | CHAR_INITIALS = CHAR_LISTS[INITIAL]
 33 | CHAR_MEDIALS = CHAR_LISTS[MEDIAL]
 34 | CHAR_FINALS = CHAR_LISTS[FINAL]
 35 | CHAR_SETS = {k: set(v) for k, v in CHAR_LISTS.items()}
 36 | CHARSET = set(itertools.chain(*CHAR_SETS.values()))
 37 | CHAR_INDICES = {k: {c: i for i, c in enumerate(v)}
 38 |                 for k, v in CHAR_LISTS.items()}
 39 | 
 40 | 
 41 | def is_hangul_syllable(c):
 42 |     return 0xac00 <= ord(c) <= 0xd7a3  # Hangul Syllables
 43 | 
 44 | 
 45 | def is_hangul_jamo(c):
 46 |     return 0x1100 <= ord(c) <= 0x11ff  # Hangul Jamo
 47 | 
 48 | 
 49 | def is_hangul_compat_jamo(c):
 50 |     return 0x3130 <= ord(c) <= 0x318f  # Hangul Compatibility Jamo
 51 | 
 52 | 
 53 | def is_hangul_jamo_exta(c):
 54 |     return 0xa960 <= ord(c) <= 0xa97f  # Hangul Jamo Extended-A
 55 | 
 56 | 
 57 | def is_hangul_jamo_extb(c):
 58 |     return 0xd7b0 <= ord(c) <= 0xd7ff  # Hangul Jamo Extended-B
 59 | 
 60 | 
 61 | def is_hangul(c):
 62 |     return (is_hangul_syllable(c) or
 63 |             is_hangul_jamo(c) or
 64 |             is_hangul_compat_jamo(c) or
 65 |             is_hangul_jamo_exta(c) or
 66 |             is_hangul_jamo_extb(c))
 67 | 
 68 | 
 69 | def is_supported_hangul(c):
 70 |     return is_hangul_syllable(c) or is_hangul_compat_jamo(c)
 71 | 
 72 | 
 73 | def check_hangul(c, jamo_only=False):
 74 |     if not ((jamo_only or is_hangul_compat_jamo(c)) or is_supported_hangul(c)):
 75 |         raise ValueError(f"'{c}' is not a supported hangul character. "
 76 |                          f"'Hangul Syllables' (0xac00 ~ 0xd7a3) and "
 77 |                          f"'Hangul Compatibility Jamos' (0x3130 ~ 0x318f) are "
 78 |                          f"supported at the moment.")
 79 | 
 80 | 
 81 | def get_jamo_type(c):
 82 |     check_hangul(c)
 83 |     assert is_hangul_compat_jamo(c), f"not a jamo: {ord(c):x}"
 84 |     return sum(t for t, s in CHAR_SETS.items() if c in s)
 85 | 
 86 | 
 87 | def split_syllable_char(c):
 88 |     """
 89 |     Splits a given korean syllable into its components. Each component is
 90 |     represented by Unicode in 'Hangul Compatibility Jamo' range.
 91 | 
 92 |     Arguments:
 93 |         c: A Korean character.
 94 | 
 95 |     Returns:
 96 |         A triple (initial, medial, final) of Hangul Compatibility Jamos.
 97 |         If no jamo corresponds to a position, `None` is returned there.
 98 | 
 99 |     Example:
100 |         >>> split_syllable_char("안")
101 |         ("ㅇ", "ㅏ", "ㄴ")
102 |         >>> split_syllable_char("고")
103 |         ("ㄱ", "ㅗ", None)
104 |         >>> split_syllable_char("ㅗ")
105 |         (None, "ㅗ", None)
106 |         >>> split_syllable_char("ㅇ")
107 |         ("ㅇ", None, None)
108 |     """
109 |     check_hangul(c)
110 |     if len(c) != 1:
111 |         raise ValueError("Input string must have exactly one character.")
112 | 
113 |     init, med, final = None, None, None
114 |     if is_hangul_syllable(c):
115 |         offset = ord(c) - 0xac00
116 |         x = (offset - offset % 28) // 28
117 |         init, med, final = x // 21, x % 21, offset % 28
118 |         if not final:
119 |             final = None
120 |         else:
121 |             final -= 1
122 |     else:
123 |         pos = get_jamo_type(c)
124 |         if pos & INITIAL == INITIAL:
125 |             pos = INITIAL
126 |         elif pos & MEDIAL == MEDIAL:
127 |             pos = MEDIAL
128 |         elif pos & FINAL == FINAL:
129 |             pos = FINAL
130 |         idx = CHAR_INDICES[pos][c]
131 |         if pos == INITIAL:
132 |             init = idx
133 |         elif pos == MEDIAL:
134 |             med = idx
135 |         else:
136 |             final = idx
137 |     return tuple(CHAR_LISTS[pos][idx] if idx is not None else None
138 |                  for pos, idx in
139 |                  zip([INITIAL, MEDIAL, FINAL], [init, med, final]))
140 | 
141 | 
142 | def split_syllables(s, ignore_err=True, pad=None):
143 |     """
144 |     Performs syllable-split on a string.
145 | 
146 |     Arguments:
147 |         s (str): A string (possibly mixed with non-Hangul characters).
148 |         ignore_err (bool): If set False, it ensures that all characters in
149 |             the string are Hangul-splittable and throws a ValueError otherwise.
150 |             (default: True)
151 |         pad (str): Pad empty jamo positions (initial, medial, or final) with
152 |             `pad` character. This is useful for cases where fixed-length
153 |             strings are needed. (default: None)
154 | 
155 |     Returns:
156 |         Hangul-split string
157 | 
158 |     Example:
159 |         >>> split_syllables("안녕하세요")
160 |         "ㅇㅏㄴㄴㅕㅇㅎㅏㅅㅔㅇㅛ"
161 |         >>> split_syllables("안녕하세요~~", ignore_err=False)
162 |         ValueError: encountered an unsupported character: ~ (0x7e)
163 |         >>> split_syllables("안녕하세요ㅛ", pad="x")
164 |         'ㅇㅏㄴㄴㅕㅇㅎㅏxㅅㅔxㅇㅛxxㅛx'
165 |     """
166 | 
167 |     def try_split(c):
168 |         try:
169 |             return split_syllable_char(c)
170 |         except ValueError:
171 |             if ignore_err:
172 |                 return (c,)
173 |             raise ValueError(f"encountered an unsupported character: "
174 |                              f"{c} (0x{ord(c):x})")
175 | 
176 |     s = map(try_split, s)
177 |     if pad is not None:
178 |         tuples = map(lambda x: tuple(pad if y is None else y for y in x), s)
179 |     else:
180 |         tuples = map(lambda x: filter(None, x), s)
181 |     return "".join(itertools.chain(*tuples))
182 | 
183 | 
184 | def join_jamos_char(init, med, final=None):
185 |     """
186 |     Combines jamos into a single syllable.
187 | 
188 |     Arguments:
189 |         init (str): Initial jao.
190 |         med (str): Medial jamo.
191 |         final (str): Final jamo. If not supplied, the final syllable is made
192 |             without the final. (default: None)
193 | 
194 |     Returns:
195 |         A Korean syllable.
196 |     """
197 |     chars = (init, med, final)
198 |     for c in filter(None, chars):
199 |         check_hangul(c, jamo_only=True)
200 | 
201 |     idx = tuple(CHAR_INDICES[pos][c] if c is not None else c
202 |                 for pos, c in zip((INITIAL, MEDIAL, FINAL), chars))
203 |     init_idx, med_idx, final_idx = idx
204 |     # final index must be shifted once as
205 |     # final index with 0 points to syllables without final
206 |     final_idx = 0 if final_idx is None else final_idx + 1
207 |     return chr(0xac00 + 28 * 21 * init_idx + 28 * med_idx + final_idx)
208 | 
209 | 
210 | def join_jamos(s, ignore_err=True):
211 |     """
212 |     Combines a sequence of jamos to produce a sequence of syllables.
213 | 
214 |     Arguments:
215 |         s (str): A string (possible mixed with non-jamo characters).
216 |         ignore_err (bool): If set False, it will ensure that all characters
217 |             will be consumed for the making of syllables. It will throw a
218 |             ValueError when it fails to do so. (default: True)
219 | 
220 |     Returns:
221 |         A string
222 | 
223 |     Example:
224 |         >>> join_jamos("ㅇㅏㄴㄴㅕㅇㅎㅏㅅㅔㅇㅛ")
225 |         "안녕하세요"
226 |         >>> join_jamos("ㅇㅏㄴㄴㄴㅕㅇㅎㅏㅅㅔㅇㅛ")
227 |         "안ㄴ녕하세요"
228 |         >>> join_jamos()
229 |     """
230 |     last_t = 0
231 |     queue = []
232 |     new_string = ""
233 | 
234 |     def flush(n=0):
235 |         new_queue = []
236 |         while len(queue) > n:
237 |             new_queue.append(queue.pop())
238 |         if len(new_queue) == 1:
239 |             if not ignore_err:
240 |                 raise ValueError(f"invalid jamo character: {new_queue[0]}")
241 |             result = new_queue[0]
242 |         elif len(new_queue) >= 2:
243 |             try:
244 |                 result = join_jamos_char(*new_queue)
245 |             except (ValueError, KeyError):
246 |                 # Invalid jamo combination
247 |                 if not ignore_err:
248 |                     raise ValueError(f"invalid jamo characters: {new_queue}")
249 |                 result = "".join(new_queue)
250 |         else:
251 |             result = None
252 |         return result
253 | 
254 |     for c in s:
255 |         if c not in CHARSET:
256 |             if queue:
257 |                 new_c = flush() + c
258 |             else:
259 |                 new_c = c
260 |             last_t = 0
261 |         else:
262 |             t = get_jamo_type(c)
263 |             new_c = None
264 |             if t & FINAL == FINAL:
265 |                 if not (last_t == MEDIAL):
266 |                     new_c = flush()
267 |             elif t == INITIAL:
268 |                 new_c = flush()
269 |             elif t == MEDIAL:
270 |                 if last_t & INITIAL == INITIAL:
271 |                     new_c = flush(1)
272 |                 else:
273 |                     new_c = flush()
274 |             last_t = t
275 |             queue.insert(0, c)
276 |         if new_c:
277 |             new_string += new_c
278 |     if queue:
279 |         new_string += flush()
280 |     return new_string


--------------------------------------------------------------------------------
/Sign_Language_Translation/video_test_model_tflite.py:
--------------------------------------------------------------------------------
  1 | import sys
  2 | import cv2
  3 | import mediapipe as mp
  4 | import numpy as np
  5 | import tensorflow as tf
  6 | import modules.holistic_module as hm
  7 | from tensorflow.keras.models import load_model
  8 | import math
  9 | import os
 10 | from PIL import ImageFont, ImageDraw, Image
 11 | from modules.utils import Vector_Normalization
 12 | fontpath = "fonts/HMKMMAG.TTF"
 13 | font = ImageFont.truetype(fontpath, 40)
 14 | 
 15 | 
 16 | actions = ['ㄱ', 'ㄴ', 'ㄷ', 'ㄹ', 'ㅁ', 'ㅂ', 'ㅅ', 'ㅇ', 'ㅈ', 'ㅊ', 'ㅋ', 'ㅌ', 'ㅍ', 'ㅎ',
 17 |              'ㅏ', 'ㅑ', 'ㅓ', 'ㅕ', 'ㅗ', 'ㅛ', 'ㅜ', 'ㅠ', 'ㅡ', 'ㅣ',
 18 |              'ㅐ', 'ㅒ', 'ㅔ', 'ㅖ', 'ㅢ', 'ㅚ', 'ㅟ']
 19 | seq_length = 10
 20 | 
 21 | # MediaPipe holistic model
 22 | detector = hm.HolisticDetector(min_detection_confidence=0.3)
 23 | 
 24 | # Load TFLite model and allocate tensors.
 25 | interpreter = tf.lite.Interpreter(model_path="models/multi_hand_gesture_classifier.tflite")
 26 | interpreter.allocate_tensors()
 27 | 
 28 | # Get input and output tensors.
 29 | input_details = interpreter.get_input_details()
 30 | output_details = interpreter.get_output_details()
 31 | 
 32 | # test video path
 33 | videoFolderPath = "dataset/example1"
 34 | videoTestList = os.listdir(videoFolderPath)
 35 | 
 36 | testTargetList =[]
 37 | 
 38 | for videoPath in videoTestList:
 39 |     actionVideoPath = f'{videoFolderPath}/{videoPath}'
 40 |     actionVideoList = os.listdir(actionVideoPath)
 41 |     for actionVideo in actionVideoList:
 42 |         fullVideoPath = f'{actionVideoPath}/{actionVideo}'
 43 |         testTargetList.append(fullVideoPath)
 44 | 
 45 | testTargetList = sorted(testTargetList, key=lambda x:x[x.find("/", 9)+1], reverse=True)
 46 | 
 47 | for target in testTargetList:
 48 | 
 49 |     cap = cv2.VideoCapture(target)
 50 | 
 51 |     seq = []
 52 |     action_seq = []
 53 |     last_action = None
 54 | 
 55 |     while cap.isOpened():
 56 |         ret, img = cap.read()
 57 |         if not ret:
 58 |             break
 59 | 
 60 |         img = detector.findHolistic(img, draw=True)
 61 |         _, right_hand_lmList = detector.findRighthandLandmark(img)
 62 | 
 63 |         if right_hand_lmList is not None:
 64 |             joint = np.zeros((42, 2))
 65 |             
 66 |             # 오른손 랜드마크 리스트
 67 |             for j, lm in enumerate(right_hand_lmList.landmark):
 68 |                 joint[j] = [lm.x, lm.y]
 69 | 
 70 | 
 71 |             # 벡터 정규화
 72 |             vector, angle_label = Vector_Normalization(joint)
 73 | 
 74 |             # 위치 종속성을 가지는 데이터 저장
 75 |             # d = np.concatenate([joint.flatten(), angle_label])
 76 |         
 77 |             # 정규화 벡터를 활용한 위치 종속성 제거
 78 |             d = np.concatenate([vector.flatten(), angle_label.flatten()])
 79 | 
 80 |             seq.append(d)
 81 | 
 82 |             if len(seq) < seq_length:
 83 |                 continue
 84 | 
 85 |             # Test model on random input data.
 86 |             # input_shape = input_details[0]['shape']
 87 |             # input_data = np.array(np.random.random_sample(input_shape), dtype=np.float32)
 88 |             
 89 |             # 시퀀스 데이터와 넘파이화
 90 |             input_data = np.expand_dims(np.array(seq[-seq_length:], dtype=np.float32), axis=0)
 91 |             input_data = np.array(input_data, dtype=np.float32)
 92 | 
 93 |             # tflite 모델을 활용한 예측
 94 |             interpreter.set_tensor(input_details[0]['index'], input_data)
 95 |             interpreter.invoke()
 96 | 
 97 |             y_pred = interpreter.get_tensor(output_details[0]['index'])
 98 |             i_pred = int(np.argmax(y_pred[0]))
 99 |             conf = y_pred[0][i_pred]
100 | 
101 |             if conf < 0.9:
102 |                 continue
103 | 
104 |             action = actions[i_pred]
105 |             action_seq.append(action)
106 | 
107 |             if len(action_seq) < 3:
108 |                 continue
109 | 
110 |             this_action = '?'
111 |             if action_seq[-1] == action_seq[-2] == action_seq[-3]:
112 |                 this_action = action
113 | 
114 |                 if last_action != this_action:
115 |                     last_action = this_action
116 |             
117 |             # 한글 폰트 출력    
118 |             img_pil = Image.fromarray(img)
119 |             draw = ImageDraw.Draw(img_pil)
120 | 
121 |             # org=(int(right_hand_lmList.landmark[0].x * img.shape[1]), int(right_hand_lmList.landmark[0].y * img.shape[0] + 20))
122 |             draw.text((10,30), f'{action.upper()}', font=font, fill=(255, 255, 255))
123 |             
124 |             img = np.array(img_pil)
125 | 
126 | 
127 |         cv2.imshow('img', img)
128 |         if cv2.waitKey(1) & 0xFF == 27:
129 |             break
130 | 
131 | 


--------------------------------------------------------------------------------
/Sign_Language_Translation/webcam_test_model_tflite.py:
--------------------------------------------------------------------------------
  1 | import sys
  2 | # sys.path.append('pingpong')
  3 | # from pingpong.pingpongthread import PingPongThread
  4 | import cv2
  5 | import mediapipe as mp
  6 | import numpy as np
  7 | import tensorflow as tf
  8 | import modules.holistic_module as hm
  9 | from tensorflow.keras.models import load_model
 10 | import math
 11 | from modules.utils import Vector_Normalization
 12 | from PIL import ImageFont, ImageDraw, Image
 13 | # from unicode import join_jamos
 14 | 
 15 | fontpath = "fonts/HMKMMAG.TTF"
 16 | font = ImageFont.truetype(fontpath, 40)
 17 | 
 18 | actions = ['ㄱ', 'ㄴ', 'ㄷ', 'ㄹ', 'ㅁ', 'ㅂ', 'ㅅ', 'ㅇ', 'ㅈ', 'ㅊ', 'ㅋ', 'ㅌ', 'ㅍ', 'ㅎ',
 19 |              'ㅏ', 'ㅑ', 'ㅓ', 'ㅕ', 'ㅗ', 'ㅛ', 'ㅜ', 'ㅠ', 'ㅡ', 'ㅣ',
 20 |              'ㅐ', 'ㅒ', 'ㅔ', 'ㅖ', 'ㅢ', 'ㅚ', 'ㅟ']
 21 | seq_length = 10
 22 | 
 23 | # MediaPipe holistic model
 24 | detector = hm.HolisticDetector(min_detection_confidence=0.3)
 25 | 
 26 | # Load TFLite model and allocate tensors.
 27 | interpreter = tf.lite.Interpreter(model_path="models/multi_hand_gesture_classifier.tflite")
 28 | interpreter.allocate_tensors()
 29 | 
 30 | # Get input and output tensors.
 31 | input_details = interpreter.get_input_details()
 32 | output_details = interpreter.get_output_details()
 33 | 
 34 | cap = cv2.VideoCapture(0)
 35 | 
 36 | seq = []
 37 | action_seq = []
 38 | last_action = None
 39 | 
 40 | # zamo_list=[]
 41 | 
 42 | while cap.isOpened():
 43 |     ret, img = cap.read()
 44 |     if not ret:
 45 |         break
 46 | 
 47 |     img = detector.findHolistic(img, draw=True)
 48 |     # _, left_hand_lmList = detector.findLefthandLandmark(img)
 49 |     _, right_hand_lmList = detector.findRighthandLandmark(img)
 50 | 
 51 |     # if left_hand_lmList is not None and right_hand_lmList is not None:
 52 |     if right_hand_lmList is not None:
 53 | 
 54 |         joint = np.zeros((42, 2))
 55 |         # 왼손 랜드마크 리스트
 56 |         # for j, lm in enumerate(left_hand_lmList.landmark):
 57 |             # joint[j] = [lm.x, lm.y]
 58 |         
 59 |         # 오른손 랜드마크 리스트
 60 |         for j, lm in enumerate(right_hand_lmList.landmark):
 61 |             # joint[j+21] = [lm.x, lm.y]
 62 |             joint[j] = [lm.x, lm.y]
 63 | 
 64 |         # 좌표 정규화
 65 |         # full_scale = Coordinate_Normalization(joint)
 66 | 
 67 |         # 벡터 정규화
 68 |         vector, angle_label = Vector_Normalization(joint)
 69 | 
 70 |         # 위치 종속성을 가지는 데이터 저장
 71 |         # d = np.concatenate([joint.flatten(), angle_label])
 72 |     
 73 |         # 벡터 정규화를 활용한 위치 종속성 제거
 74 |         d = np.concatenate([vector.flatten(), angle_label.flatten()])
 75 | 
 76 |         # 정규화 좌표를 활용한 위치 종속성 제거 
 77 |         # d = np.concatenate([full_scale, angle_label.flatten()])
 78 |         
 79 | 
 80 |         seq.append(d)
 81 | 
 82 |         if len(seq) < seq_length:
 83 |             continue
 84 | 
 85 |         # Test model on random input data.
 86 |         # input_shape = input_details[0]['shape']
 87 |         # input_data = np.array(np.random.random_sample(input_shape), dtype=np.float32)
 88 |         
 89 |         # 시퀀스 데이터와 넘파이화
 90 |         input_data = np.expand_dims(np.array(seq[-seq_length:], dtype=np.float32), axis=0)
 91 |         input_data = np.array(input_data, dtype=np.float32)
 92 | 
 93 |         # tflite 모델을 활용한 예측
 94 |         interpreter.set_tensor(input_details[0]['index'], input_data)
 95 |         interpreter.invoke()
 96 | 
 97 |         y_pred = interpreter.get_tensor(output_details[0]['index'])
 98 |         i_pred = int(np.argmax(y_pred[0]))
 99 |         conf = y_pred[0][i_pred]
100 | 
101 |         if conf < 0.9:
102 |             continue
103 | 
104 |         action = actions[i_pred]
105 |         action_seq.append(action)
106 | 
107 |         if len(action_seq) < 3:
108 |             continue
109 | 
110 |         this_action = '?'
111 |         if action_seq[-1] == action_seq[-2] == action_seq[-3]:
112 |             this_action = action
113 | 
114 |             if last_action != this_action:
115 |                 last_action = this_action
116 |         '''
117 |         # 기록된 한글 파악
118 |         if zamo_list[-1] != this_action: # 만약 전에 기록된 글자와 이번 글자가 다르다면
119 |             zamo_list.append(this_action)
120 |         
121 |         zamo_str = ''.join(zamo_list) # 리스트에 있는 단어 합침
122 |         unitl_action = join_jamos(zamo_str) # 합친 단어 한글로 만들기
123 |         '''
124 |         
125 |         # 한글 폰트 출력    
126 |         img_pil = Image.fromarray(img)
127 |         draw = ImageDraw.Draw(img_pil)
128 |         '''
129 |         draw.text((int(right_hand_lmList.landmark[0].x * img.shape[1]), int(right_hand_lmList.landmark[0].y * img.shape[0] + 20)),
130 |                   f'{this_action.upper()}', 
131 |                   font=font, 
132 |                   fill=(255, 255, 255))
133 |         '''
134 |         draw.text((10, 30), f'{action.upper()}', font=font, fill=(255, 255, 255))
135 | 
136 |         img = np.array(img_pil)
137 | 
138 |         
139 |         
140 |         
141 |         # cv2.putText(img, f'{this_action.upper()}', org=(int(right_hand_lmList.landmark[0].x * img.shape[1]), int(right_hand_lmList.landmark[0].y * img.shape[0] + 20)), fontFace=cv2.FONT_HERSHEY_SIMPLEX, fontScale=1, color=(255, 255, 255), thickness=2)
142 | 
143 | 
144 |     cv2.imshow('img', img)
145 |     if cv2.waitKey(1) & 0xFF == 27:
146 |         break
147 | 
148 | 


--------------------------------------------------------------------------------
/dataset/output_video/ㄱ/ㄱ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㄱ/ㄱ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㄴ/ㄴ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㄴ/ㄴ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㄷ/ㄷ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㄷ/ㄷ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㄹ/ㄹ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㄹ/ㄹ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅁ/ㅁ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅁ/ㅁ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅂ/ㅂ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅂ/ㅂ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅅ/ㅅ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅅ/ㅅ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅇ/ㅇ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅇ/ㅇ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅈ/ㅈ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅈ/ㅈ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅊ/ㅊ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅊ/ㅊ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅋ/ㅋ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅋ/ㅋ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅌ/ㅌ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅌ/ㅌ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅍ/ㅍ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅍ/ㅍ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅎ/ㅎ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅎ/ㅎ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅏ/ㅏ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅏ/ㅏ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅐ/ㅐ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅐ/ㅐ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅑ/ㅑ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅑ/ㅑ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅒ/ㅒ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅒ/ㅒ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅓ/ㅓ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅓ/ㅓ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅔ/ㅔ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅔ/ㅔ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅕ/ㅕ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅕ/ㅕ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅖ/ㅖ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅖ/ㅖ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅗ/ㅗ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅗ/ㅗ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅚ/ㅚ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅚ/ㅚ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅛ/ㅛ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅛ/ㅛ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅜ/ㅜ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅜ/ㅜ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅟ/ㅟ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅟ/ㅟ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅠ/ㅠ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅠ/ㅠ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅡ/ㅡ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅡ/ㅡ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅢ/ㅢ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅢ/ㅢ_1.avi


--------------------------------------------------------------------------------
/dataset/output_video/ㅣ/ㅣ_1.avi:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/output_video/ㅣ/ㅣ_1.avi


--------------------------------------------------------------------------------
/dataset/seq_ㄱ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄱ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㄱ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄱ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㄱ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄱ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㄴ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄴ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㄴ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄴ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㄴ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄴ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㄷ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄷ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㄷ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄷ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㄷ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄷ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㄹ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄹ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㄹ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄹ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㄹ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㄹ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅁ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅁ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅁ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅁ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅁ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅁ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅂ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅂ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅂ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅂ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅂ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅂ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅅ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅅ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅅ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅅ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅅ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅅ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅇ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅇ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅇ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅇ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅇ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅇ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅈ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅈ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅈ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅈ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅈ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅈ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅊ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅊ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅊ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅊ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅊ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅊ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅋ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅋ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅋ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅋ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅋ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅋ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅌ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅌ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅌ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅌ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅌ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅌ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅍ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅍ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅍ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅍ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅍ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅍ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅎ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅎ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅎ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅎ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅎ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅎ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅏ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅏ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅏ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅏ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅏ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅏ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅐ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅐ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅐ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅐ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅐ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅐ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅑ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅑ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅑ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅑ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅑ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅑ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅒ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅒ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅒ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅒ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅒ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅒ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅓ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅓ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅓ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅓ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅓ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅓ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅔ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅔ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅔ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅔ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅔ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅔ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅕ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅕ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅕ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅕ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅕ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅕ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅖ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅖ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅖ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅖ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅖ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅖ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅗ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅗ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅗ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅗ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅗ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅗ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅚ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅚ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅚ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅚ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅚ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅚ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅛ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅛ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅛ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅛ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅛ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅛ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅜ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅜ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅜ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅜ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅜ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅜ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅟ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅟ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅟ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅟ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅟ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅟ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅠ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅠ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅠ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅠ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅠ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅠ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅡ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅡ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅡ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅡ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅡ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅡ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅢ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅢ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅢ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅢ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅢ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅢ_1669724266.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅣ_1669720403.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅣ_1669720403.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅣ_1669723415.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅣ_1669723415.npy


--------------------------------------------------------------------------------
/dataset/seq_ㅣ_1669724266.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/dataset/seq_ㅣ_1669724266.npy


--------------------------------------------------------------------------------
/models/multi_hand_gesture_classifier.h5:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/models/multi_hand_gesture_classifier.h5


--------------------------------------------------------------------------------
/models/multi_hand_gesture_classifier.tflite:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/models/multi_hand_gesture_classifier.tflite


--------------------------------------------------------------------------------
/readme.md:
--------------------------------------------------------------------------------
 1 | ## ***실시간 수어 번역 인식 모듈생성 (with. MediaPipe, LSTM)***
 2 | ***
 3 | ### Team Project
 4 | 
 5 | #### Teaming   최준용, 이태범, 서동혁 ***(in Kookmin_University)***
 6 | 
 7 | ***
 8 | 
 9 | - 실시간 영상 및 녹화 동영상을 활용해 지문자 인식 프로그램 생성
10 | 
11 | - 활용방안
12 |     - 수어 사용자를 위한 학습 보조 프로그램 개발 가능
13 |     - 실시간 수어 사용자와의 의사소통 프로그램 개발 가능
14 | 
15 | - 수어종류
16 |     - 자모음(31개에 대한 한글 자모 지문자)
17 |     
18 |     
19 | ![지문자_이미지](https://user-images.githubusercontent.com/90700892/209419165-fd373820-e70c-4a1b-b2a4-c82439db4c1c.jpg)
20 | 
21 | 출처 : https://www.urimal.org/1222
22 | 
23 | ***
24 | 
25 | ### ***데이터 수집***
26 | 
27 | - 31개의 자음, 모음에 대한 팀원 3명의 학습영상 촬영
28 | 
29 | ![팀원3명데이터생성](https://user-images.githubusercontent.com/90700892/209419274-abda09a1-fd5b-4a8e-b018-2c04209db293.gif)
30 | 
31 | 각각 'ㅎ' , 'ㅏ' , 'ㄱ' 에 대한 data 생성중...
32 | 
33 | ***
34 | 
35 | ### ***데이터 전처리***
36 | 
37 | ![hand_landmarks](https://user-images.githubusercontent.com/90700892/209419270-aad3fcde-48b7-40cf-b2f6-a38edeee1e89.png)
38 | 
39 | 출처 : https://google.github.io/mediapipe/solutions/hands.html
40 | 
41 | 수집한 영상 데이터에서 위와 같이 각 hand keypoint의 Vector, Angle 값을 인식해 데이터로 사용하게 됩니다.
42 | 
43 | 본 프로젝트에서는 지문자를 활용하기 때문에 한손 keypoint만 활용하였습니다.
44 | 
45 | ***
46 | 
47 | ### ***Pipeline***
48 | 
49 | - making_video.py
50 |     - 원하는 자,모음을 설정해 동영상을 생성합니다. (openCV 활용)
51 |     
52 | - create_dataset_from_video.py
53 |     - video data를 사용하여 hand keypoint의 Vector, Angle 값을 sequence data로 변환해 npy 파일로 저장합니다.
54 |     
55 | - train_hand_gesture.ipynb
56 |     - npy file load하여 모델을 생성합니다.
57 |     
58 | - video_test_model_tflite.py
59 |     - videoFolderPath를 지정하여 저장된 비디오를 활용하여 테스트합니다.
60 |     
61 | - webcam_test_model_tflite.py
62 |     - webcam을 활용하여 실시간으로 테스트합니다.
63 | 
64 | ***
65 | 
66 | ### ***시연***
67 | 
68 | #### Using webcam
69 | 
70 | ![KakaoTalk_20221210_214325078](https://user-images.githubusercontent.com/90700892/209419238-00fee2f8-179d-4632-9e84-a1c18cf81b94.gif)
71 | 
72 | ***
73 | 
74 | ### ***결론***
75 | 
76 | 저작권 문제로 여기엔 넣지 못했지만 타 youtube 동영상에 대입시켜본 결과 정확히 지문자를 분류 하는 성능을 보였습니다.  
77 | 추가적으로 왼쪽 위에 뜨는 문자들을 결합해 text data까지 생성시킬 수 있도록 project를 향후 발전 시킬 계획에 있습니다.
78 | 


--------------------------------------------------------------------------------
/version_requirements.txt:
--------------------------------------------------------------------------------
1 | tensorflow==2.4.0
2 | sklearn
3 | numpy
4 | opencv-python
5 | mediapipe


--------------------------------------------------------------------------------
/실시간 수화 번역 인식 모듈 생성.pptx:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/JunYong-Choi/Sign_Language_Translation/d3374ddcc6b5532e683db2275e7ad2074dc9627d/실시간 수화 번역 인식 모듈 생성.pptx


--------------------------------------------------------------------------------