Trained with BDD dataset or COCO?

zhanghanduo / yolo3_pytorch

Yolo3 Implementation in Pytorch using COCO and BDD100K dataset. Forked from https://github.com/BobLiu20/YOLOv3_PyTorch.git. For my own fun, so the introduction is not accurate.

104 stars 21 forks source link

Trained with BDD dataset or COCO? #7

Closed TheIllusion closed 4 years ago

TheIllusion commented 4 years ago

Hi, I'm impressed by this repo and I'm using the pretrained models with joy.

However, I'm wondering if the pretrained models are trained with BDD or COCO.

Could you give me an answer?

zhanghanduo commented 4 years ago

Hi. I didn't provide the pretrained models with BDD. It is still COCO. This repo was was developed on top of another popular repo and I simply did some changes. When I was training with BDD, I met some issues and the accuracy was not satisfactory. Then I was busy doing something else and you see I stopped updating code a long time ago. Sorry if this cannot help you.

Currently there are many new state-of-the-art object detection worth trying instead of YOLOv3. I recommend centernet. Good luck.

TheIllusion commented 4 years ago

Thanks for the reply.