search

hhk7734 / tensorflow-yolov4

YOLOv4 Implemented in Tensorflow 2.
MIT License
136 stars 75 forks source link

No Boxes #91

Open Sayyam-Jain opened 3 years ago

Sayyam-Jain commented 3 years ago

Hi, thank you for the wonderful work. I tried it on default tiny yolov4 weights and used yolov4-tiny-relu-tpu.cfg for conversion. I also used coco validation dataset for post-training quantization. After that, I ran edgecompiler and the weights are converted successfully.

However, when I run inference, it fails to detect any box. Can you please help? The conversion and inference script are same as in your wiki link.

Thanks

AntoAndGar commented 2 years ago

I have the same problem with custom weights

hhk7734 commented 2 years ago

My weight file is not good. If you want to get a more acurracy model. train data with darknet. the configuration file is compatible with darkent.