All of this code is adapted/ported/copied from https://github.com/rwightman/efficientdet-pytorch/tree/2fc4e0e107c969090380d0be1a7ad7552dc00c63
Modification List:
- Remove unused files and dataset.
- Adding Bus Dataset (Base on pickle).
- Remove support of distributed mode.
- Adding logging via weights and biases
As first step, you should built the dataset one of the scripts in .scripts
- Regression Loss
- GIoU
- CIoU
- Mean-Variance estimation
- Replace BN - change BN due to small batch size on single GPU.
- Data Augementation
- Mosic
- CutPastObject