Easy training on custom dataset. Various backends (MobileNet and SqueezeNet) supported. A YOLO demo to detect raccoon run entirely in brower is accessible at https://git.io/vF7vI (not on Windows).
At first ,Thank you for your awesome work.
I learn that the pretrain weights is trained on the COCO2014 dataset and I'm a little confused,some quenstions are listed:
The Pretrained weights is all the weights include the backend and frontend ?Could i use the classfication model pretrain weights only for the feature extraction?
If my own dataset is quite different from the COCO2014 dataset ,what should i do to improve the model?
At first ,Thank you for your awesome work. I learn that the pretrain weights is trained on the COCO2014 dataset and I'm a little confused,some quenstions are listed:
The Pretrained weights is all the weights include the backend and frontend ?Could i use the classfication model pretrain weights only for the feature extraction?
If my own dataset is quite different from the COCO2014 dataset ,what should i do to improve the model?
thank you very much!