search

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 784 forks source link

Problem on download pretrained weight. #264

Open HanYuanyuaner opened 6 years ago

HanYuanyuaner commented 6 years ago

I can not download the pretrained weight from https://1drv.ms/f/s!ApLdDEW3ut5fec2OzK4S4RpT-SU or https://1drv.ms/f/s!ApLdDEW3ut5feoZAEUwmSMYdPlY. Can anyone share it ?

experiencor commented 6 years ago

Use proxysite.com

houxy12 commented 6 years ago

proxysite.com doesn't work for me. @experiencor

HanYuanyuaner commented 6 years ago

It also doesn't work for me

GitMsw commented 6 years ago

Can share another one? in need

lychenpan commented 6 years ago

I have transformed one yolov2.weights(without the last linear layer) to h5 file using this script, you can download it here

mdawas commented 5 years ago

I have the same problem, I can't download the pretrained weights for the whole model and I didn't find the Full_Yolo weights too! Anyone got that problem solved?

rodrigo2019 commented 5 years ago

https://github.com/rodrigo2019/keras-yolo2/releases

mdawas commented 5 years ago

@rodrigo2019 Thanks alot! it worked!

mdawas commented 5 years ago

@rodrigo2019 I have a problem when using the Full Yolo backend with the weights in the mentioned repo, the training can't start and the following error is logged:

Traceback (most recent call last): File "train.py", line 101, in main(args) File "train.py", line 70, in main anchors = config['model']['anchors']) File "/geo/home/i501245/projects/keras-yolo/keras-yolo2/frontend.py", line 48, in init self.feature_extractor = FullYoloFeature(self.input_size) File "/geo/home/i501245/projects/keras-yolo/keras-yolo2/backend.py", line 167, in init self.feature_extractor.load_weights(FULL_YOLO_BACKEND_PATH) File "/geo/home/d067913/envs/miniconda2/envs/mohammad/lib/python2.7/site-packages/keras/engine/topology.py", line 2645, in load_weights load_weights_from_hdf5_group(f, self.layers) File "/geo/home/d067913/envs/miniconda2/envs/mohammad/lib/python2.7/site-packages/keras/engine/topology.py", line 3139, in load_weights_from_hdf5_group str(len(filtered_layers)) + ' layers.') ValueError: You are trying to load a weight file containing 45 layers into a model with 44 layers.

Have you got the same error or there is something should be modified in building the network?