The model brings several changes with some paradigms:
It only allows train, test and val dataset
The val dataset is used to perform evaluations at the end of each epoch
The test dataset is used to perform the evaluation at the end of the training
Everything located inside yolox-detection/experiment/YOLOX SHOULD NOT be review, as it comes directly from the original yolox repository.
[x] Only feature that remains to be developed is handling 1 or 2 datasets (for now only a training with exactly 3 datasets is supported)
-> Warning ⚠️ : Some files have been linted by black during the development phase, another PR that integrate black as a pre-commit step should be included
This PR brings the new yolox model that comes from here: https://github.com/Megvii-BaseDetection/YOLOX.
The model brings several changes with some paradigms:
train,testandvaldatasetvaldataset is used to perform evaluations at the end of each epochtestdataset is used to perform the evaluation at the end of the trainingyolox-detection/experiment/YOLOXSHOULD NOT be review, as it comes directly from the original yolox repository.blackduring the development phase, another PR that integrateblackas a pre-commit step should be included