Hi lzx,
I'd like to contribute four SSD/RFBNet models with HarDNet68/85 as backbone models to this repo.
Please note that:
utils/box_utils.py: I believe there is a bug with the fully reduced dim for the sum
I also made some modifications for supporting pytorch 0.4.1 ~ 1.1.0. I've tested it with 0.4.1, 1.0.1, and 1.1.0. However, I couldn't test it with the earlier versions due to the cuda version, so I'm not sure if it will conflict to 0.3.1 and earlier versions.
If there is any concern with the modification on train_test.py, we can rename it to train_test_hardnet.py
Thank you so much!
Hi lzx, I'd like to contribute four SSD/RFBNet models with HarDNet68/85 as backbone models to this repo. Please note that: