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 787 forks source link

Link to the pretrained weights for the whole model #393

Open amapic opened 5 years ago

amapic commented 5 years ago

The link given in the README is not working. How to have those weights ?

Bergo1305 commented 5 years ago

You can use yolov2 weights from darknet, but firstly you need to convert into keras model.

amapic commented 5 years ago

Where can I find weights from darknet ? Sorry I am a beginner. It's not easy when you discover Linux in the same time... haha

zoecarver commented 5 years ago

@rodrigo2019 has some pretrained waits which work well here.

Noam-M commented 5 years ago

@zoecarver have you tried using these weights? I tried and I'm getting the following error: ValueError: Cannot feed value of shape (425, 1024, 1, 1) for Tensor 'Placeholder_110:0', which has shape '(1, 1, 1024, 55)'

I'm using a Full Yolo model of input size 416

kashyap-dhriti commented 5 years ago

https://gitlab.com/ml-tools/yolo-pretrained-weights download yolo weights from here.

GANWANSHUI commented 5 years ago

does anyone get the usable pretrained weight?the link does'n work,so sad!

lijingpeng commented 5 years ago

So sad all the link about pretrained weight just do not work at all.

ValueError: Cannot feed value of shape (425, 1024, 1, 1) for Tensor u'Placeholder_110:0', which has shape '(1, 1, 1024, 55)'

It seems like the pretrained weight is irrelevant with this model here

MuhammadAsadJaved commented 4 years ago

@lijingpeng @GANWANSHUI Have you found the pre-trained weights for the whole model?