DL-semantic-segmentation

Data Files

Dataset Contents

This dataset version consists of 10K images with annotations for 7 tasks.

• RGB images
• Semantic segmentation
• 2D bounding boxes
• Instance segmentation
• Motion segmentation
• Previous images
• CAN information
• Lens soiling data and annotations
• Calibration Information
• Dense polygon points for objects

Coming Soon:

• Fisheye sythetic data with semantic annotations
• Lidar and dGPS scenes

To download data or to learn more details about the data, please go here and here.

The whole picture of files

• ./privous_images/ contains 8234 png files.

• ./rgb_images/ contains 8234 png files.

• ./rgb_images(test_set)/ contains 1766 png files.

• ./train/:

  • ./train/gtLabels/ contains 5000 png files (black).
  • ./train/rgbImages/ contains 5000 png files (normal).
  • ./train/rgbLabels/ contains 5000 png files (RGB labels).

• ./test/:

  • ./test/gtLabels/ contains 1000 png files (black).
  • ./test/rgbImages/ contains 1000 png files (normal).
  • ./test/rgbLabels/ contains 1000 png files (RGB labels).

• ./WoodScape_ICCV19/:

  Click me

    - `~/box_2d_annotations/`:
  
        - `box_2d_annotation_info.json`
  
        - `~/box_2d_annotations/` contains 8234 txt files.
  
    - `~/calibration_data/`:
  
        - `calibration_readme.md`
  
        - `~/calibration/` contains 8234 json files.
  
    - `~/instance_annotations/`:
  
        - `~/class_info.json`
  
        - `~/instance_annotations/` contains 8234 json files.
  
    - `~/motion_annotations/`:
  
        - `~/motion_annotation_info.json`
  
        - `~/motion_annotations/`:
  
            - `~/gtLabels/` contains 8234 png files.
  
            - `~/rgbLabels` contains 8234 png files.
  
    - `~/semantic_annotations/`:
  
        - `seg_annotation_info.json`
  
        - `~/semantic_annotations/`:
  
            - `~/gtLabels/` contains 8234 png files.
  
            - `~/rgbLabels` contains 8234 png files.
  
    - `~/soiling_dataset/`:
  
        - `~/soiling_annotation_info.json`
  
    - `~/vehicle_data/`:
  
        - `~/vehicle_info/` contains 8234 json files.
  

To download data or to learn more details about the data, please go here and here.

Data organization

Click me to see details

woodscape
│   README.md    
│
└───rgb_images
│   │   00001_[CAM].png
│   │   00002_[CAM].png
|   |   ...
│   │
└───previous_images
│   │   00001_[CAM]_prev.png
│   │   00002_[CAM]_prev.png
|   |   ...
│   │
└───semantic_annotations
        │   rgbLabels
        │   │   00001_[CAM].png
        │   │   00002_[CAM].png
        |   |   ...
        │   gtLabels
        │   │   00001_[CAM].png
        │   │   00002_[CAM].png
        |   |   ...
│   │
└───box_2d_annotations
│   │   00001_[CAM].png
│   │   00002_[CAM].png
|   |   ...
│   │
└───instance_annotations
│   │   00001_[CAM].json
│   │   00002_[CAM].json
|   |   ...
│   │
└───motion_annotations
        │   rgbLabels
        │   │   00001_[CAM].png
        │   │   00002_[CAM].png
        |   |   ...
        │   gtLabels
        │   │   00001_[CAM].png
        │   │   00002_[CAM].png
        |   |   ...
│   │
└───vehicle_data
│   │   00001_[CAM].json
│   │   00002_[CAM].json
|   |   ...
│   │
│   │
└───calibration_data
│   │   00001_[CAM].json
│   │   00002_[CAM].json
|   |   ...
│   │
└───soiling_dataset
        │   rgb_images
        │   │   00001_[CAM].png
        │   │   00002_[CAM].png
        |   |   ...
        │   gt_labels
        │   │   00001_[CAM].png
        │   │   00002_[CAM].png
        |   |   ...
        │   gt_labels
        │   │   00001_[CAM].png
        │   │   00002_[CAM].png
        |   |   ...

[CAM] :

• FV --> Front CAM
• RV --> Rear CAM
• MVL --> Mirror Left CAM
• MVR --> Mirror Right CAM

To download data or to learn more details about the data, please go here and here.

Code

• args.py defines the parser of arguments.

• main.py defines the main program.

• predict.py predict labels for testing images and generate output PNG files.

• utils.py defines appropriate formats of dataset and data loader and pipelines of data augmentation.

• ./smpgit/segmentation_models_pytorch/: socure codes of segmentation_models_pytorch

• visualization.ipynb: plots of metrics and visualization of predictions

Usage

python3 main.py -enc 'efficientnet-b4' -e 100 -d 'cuda:0'
python3 predict.py -Pm '16-02-58-45efficientnet-b4_bs=8_epochs=30'

Reference

1. Saravanabalagi Ramachandran, John McDonald, and Ganesh Sistu (2021). Woodscape Fisheye Semantic Segmentation for Autonomous Driving | CVPR 2021 OmniCV Workshop Challenge. https://competitions.codalab.org/competitions/30993.

2. Pavel Yakubovskiy (2020). Segmentation Models Pytorch. GitHub: GitHub repository. https://github.com/qubvel/segmentation_models.pytorch.

3. Jeremy Jordan (2018). Evaluating image segmentation models. https://www.jeremyjordan.me/evaluating-image-segmentation-models/.

4. Chen, L. C., Papandreou, G., Schroff, F., & Adam, H. (2017). Rethinking atrous convolution for semantic image segmentation. arXiv preprint arXiv:1706.05587.

5. Arunava Chakraborty (2019). PyTorch for Beginners: Semantic Segmentation using torchvision. https://learnopencv.com/pytorch-for-beginners-semantic-segmentation-using-torchvision/.

6. Saurabh Kumar (2020). Semantic hand segmentation using Pytorch. Medium: towards data science. https://towardsdatascience.com/semantic-hand-segmentation-using-pytorch-3e7a0a0386fa.