├── 71.jpeg
├── output.png
├── README.md
└── tester.py


/71.jpeg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/noman55/ImageSegmentation_AI/HEAD/71.jpeg


--------------------------------------------------------------------------------
/output.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/noman55/ImageSegmentation_AI/HEAD/output.png


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
 1 | # AutonomousCar
 2 | 
 3 | Here we are trying to use the famous technique called Semantic Segmentation. I have alredy trained the model using Fully Conneted
 4 | Convolution Network on Pascal VOC dataset which you can download from cityscapes official website.
 5 | 
 6 | Here is the input image given to modal 
 7 | 
 8 | ![](71.jpeg)
 9 | 
10 | and here is the result 
11 | 
12 | ![](output.png)
13 | 
14 | since the size of the trained model is >100MB i have not uploaded.
15 | 
16 | 


--------------------------------------------------------------------------------
/tester.py:
--------------------------------------------------------------------------------
 1 | 
 2 | import mxnet as mx
 3 | from mxnet import image
 4 | from mxnet.gluon.data.vision import transforms
 5 | import gluoncv
 6 | import cv2
 7 | # using cpu
 8 | ctx = mx.cpu(0)
 9 | 
10 | filename= '71.jpeg'
11 | # load the image
12 | img = image.imread(filename)
13 | 
14 | from matplotlib import pyplot as plt
15 | #plt.imshow(img.asnumpy())
16 | #plt.show()
17 | from gluoncv.data.transforms.presets.segmentation import test_transform
18 | img = test_transform(img, ctx)
19 | # get pre-trained model
20 | model = gluoncv.model_zoo.get_model('fcn_resnet101_voc', pretrained=True)
21 | output = model.predict(img)
22 | predict = mx.nd.squeeze(mx.nd.argmax(output, 1)).asnumpy()
23 | from gluoncv.utils.viz import get_color_pallete
24 | import matplotlib.image as mpimg
25 | mask = get_color_pallete(predict, 'pascal_voc')
26 | mask.save('output.png')
27 | 


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