why the mAP is not good(28.6%)?

experiencor / keras-yolo2

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).

MIT License

1.73k stars 785 forks source link

why the mAP is not good(28.6%)? #377

Open HHaikuo opened 5 years ago

HHaikuo commented 5 years ago

Is the model not well trained, or is it the reason for code?

MuhammadAsadJaved commented 4 years ago

@HHaikuo ho do you calculate mAP?