└── README.md


/README.md:
--------------------------------------------------------------------------------
 1 | # Thermal-NeRF
 2 | The official dataset for our paper: [**Thermal-NeRF: Neural Radiance Fields from an Infrared Camera**](https://arxiv.org/abs/2403.10340).
 3 | ## Self-collected Infrared Dataset 
 4 | ### 1.1 Description
 5 | The dataset utilizes the infrared camera for image collection, with ground truth camera poses recorded using the VICON motion capture system (MoCap) and camera intrinsics obtained through the infrared calibration. It features three distinct scenarios designed to replicate visually degraded environments, including low lighting, fluctuating lighting, and smoke. These scenarios aim to thoroughly evaluate the performance of thermal imaging under challenging conditions. 
 6 | 
 7 | ### 1.2 Sensor Parameter
 8 | * **Infrared Camera:** optris PI 450i, 382*288
 9 | * **Rostopics of rosbag sequences:** /optris/thermal_image
10 | 
11 | ### 1.3 Data Format 
12 | - The dataset can be downloaded from [this link](https://drive.google.com/drive/folders/11Xtrd09bIzCifB1hicU9E9lgKrGXFcSB?usp=sharing).
13 | 
14 | - The data structure of the dataset is shown as below: 
15 | ```
16 | /Thermal-NeRF dataset
17 | │
18 | ├── /low lighting
19 | │   ├── /sequence_1
20 | │   │   ├── /train
21 | │   │   │   ├── /intrinsics
22 | │   │   │   ├── /pose
23 | │   │   │   └── /rgb
24 | │   │   └── /test
25 | │   │       ├── /intrinsics
26 | │   │       ├── /pose
27 | │   │       └── /rgb
28 | │   ├── /sequence_2
29 | │   ...
30 | │ 
31 | ├── /fluctuating lighting
32 | │   ├── /sequence_1
33 | │   ...
34 | │
35 | ├── /smoke
36 | │   ├── /sequence_1
37 | │   ...
38 | ```
39 | - All the pose data conformes to the **OpenCV** camera convention, adaption should be made if used with the OpenGL camera convention.
40 | 


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