in the readme, I saw the tiny-yolov7 performs quite well. Did you train it with this repository or with the yolov7 repository?
Can you provide the training hyperparamters? I tried it as well, but my results were far from optimal. I think the mosaic augmentation may degrade the performance (as you have written in the paper), but without the mosaic-augmentation I get weird errors at some point during the training when using the original yolov7 repository.
Dear authors,
in the readme, I saw the tiny-yolov7 performs quite well. Did you train it with this repository or with the yolov7 repository?
Can you provide the training hyperparamters? I tried it as well, but my results were far from optimal. I think the mosaic augmentation may degrade the performance (as you have written in the paper), but without the mosaic-augmentation I get weird errors at some point during the training when using the original yolov7 repository.
Cheers, Dennis