The projects is set of tools prepared for pre- or postprocessing data in Computer Vision domain. Fow images aw well as videos. Main goal is to have codebase for repetetive tasks in many projects.
STATUS: Things to be done marked in the code.
For showcase/reference please go to DEMO notebook.
Check samples below for an idea:
Fast video annotation:
Frame from video: https://www.youtube.com/watch?v=MNn9qKG2UFI which was used during tests in ths project.
Object apperance heatmaps:
Plot of object count:
Ultra-quick baseline annotations:
The code addresses taksks like:
- image representations (PIL/OpenCV/RGB vs BRG issues)
- image cropping & scaling
- bbox notation standarization
- video heatmaps
- caching detections to pickle file
Visual helper methods include:
- bounding box drawing
Code uses Python 3 syntax.
Core code uses:
Sample are using additionally:
- PyTorch (torch, torchvision)
This code is licensed with Apache License Version 2.0.
Original author: Błażej Matuszewski, repository: bwosh/annovid.
Used libraries are licensed separately. Please review them as well.