PyTorch Implementation of Complex-YOLO 3D
Much of the code has been adapted from this repo. However, I've changed a considerable amount of code, improved readibility and speed (from initial runs).
Update:
- 3/3 - Fixed bug in loss. Commented
region_loss.pyandutils.pyin relation to notation in paper. - 3/2 - Not able understand what loss function was doing. Bug in loss.
- Go through Kitti Dataset class definition
- Construct model from original implementation
- Download data
- Modify loss function
- Comment loss function. Understand everything that's going on
- Test function
- Visualization of results
- Running loss for saving model
- Streamline code in
region_loss.py. Two for loops are unnecessary