bdd100k Visualization Tool

bdd100k_vis is a Dash webapp that provides visualizations for the Berkley Deep Drive Dataset bdd100k.

This tool is also available as a TKinter GUI application (see branches). The Dash implementation is partially based off of the DETR Detection App.

Main Features:

• Browse and view images
• Semantic/instance segmentation visualization:
  • Overlay using dynamic alpha composite (user adjustable)
• Object detection visualization:
  • Draw bounding boxes from JSON labels file
• Legend for object colors

Usage

Clone this repo, and run the python script from the repo directory:

git clone doronser/bdd100k_visualize

cd bdd100k_visualize

python bdd100k_vis.py

http://127.0.0.1:8050/

Data Setup

This repo includes a dummy dataset to demonstrante the capabilities of the tool using several images. If you want to get the entire dataset, it can downloaded here. The parts of the dataset supported by this tool are:

• 100k images
• 10k images
• Detection 2020 Labels
• Semantic Segmentation
• Instance Sementation

See bddk100k documentation for more information.

Once you have the dataset, modify bdd100k_vis.py to include the path to the dataset like so:

 app = bdd100k_vis('path/to/dataset/bdd100k')

Screenshots

Object Detection:

Instance Segmentation:

Semantic Segmentation:

License

MIT Open Source

Feel free to use this work as long as you refrence this repo.

Contact: [email protected]