Closed campos537 closed 5 years ago
I liked a lot the results you had, but i want to train using this other CNN that are faster, to run on CPU
@campos537 You can replace Res101 (L408 and L465) with new backbones and modify L228, L253-L254 and L298 correspondingly.
Thanks for the help!
Can you share your script to extract the pretrained COCO model of given classes? @sfzhang15
@wait1988 Please refer to convert_model_320.ipynb and convert_model_512.ipynb.
I liked a lot the results you had, but i want to train using this other CNN that are faster, to run on CPU