Pytorch-SiamFC icon indicating copy to clipboard operation
Pytorch-SiamFC copied to clipboard

can't find "baseline-conv5-original" network

Open arbitularov opened this issue 6 years ago • 2 comments

Here "Download the state dictionary of the Network and put it inside of Root_Dir/models and set the value of net in Root_Dir/tracking/experiments/<your_experiments>/design.json as the name of the network (e.g. baseline-conv5-original)" I don't understand how to take " baseline-conv5-original" network. I did it: "net": "baseline_pretrained.pth.tar" Then I have a mistake: xgbj@Name:/Tracker/Pytorch-SiamFC$ chmod -x run_tracker_evaluation.py xgbj@Name:/Tracker/Pytorch-SiamFC$ ./run_tracker_evaluation.py Model set to GPU

Traceback (most recent call last): File "./run_tracker_evaluation.py", line 235, in main() File "./run_tracker_evaluation.py", line 32, in main tracking_graph = TrackingGraph(design, env, hp) File "/Tracker/Pytorch-SiamFC/tracking/tracking_graph.py", line 43, in init self.siamese = load_model(self.siamese, design.net) File "/Tracker/Pytorch-SiamFC/tracking/networks.py", line 90, in load_model model.load_state_dict(torch.load(model_path)) File "/home/xgbj/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 719, in load_state_dict self.class.name, "\n\t".join(error_msgs))) RuntimeError: Error(s) in loading state_dict for SiamFC: Missing key(s) in state_dict: "branch.0.weight", "branch.0.bias", "branch.1.weight", "branch.1.bias", "branch.1.running_mean", "branch.1.running_var", "branch.4.weight", "branch.4.bias", "branch.5.weight", "branch.5.bias", "branch.5.running_mean", "branch.5.runn ing_var", "branch.8.weight", "branch.8.bias", "branch.9.weight", "branch.9.bias", "branch.9.running_mean", "branch.9.running_var", "bra nch.11.weight", "branch.11.bias", "branch.12.weight", "branch.12.bias", "branch.12.running_mean", "branch.12.running_var", "branch.14.w eight", "branch.14.bias", "match_batchnorm.weight", "match_batchnorm.bias", "match_batchnorm.running_mean", "match_batchnorm.running_va r". Unexpected key(s) in state_dict: "epoch", "state_dict", "optim_dict".

arbitularov avatar Nov 07 '18 06:11 arbitularov

You're right, there's a conflict between the Training and Tracking code. Somewhere along the development the definitions diverged. I'll be working to correct it and get back to you ASAP.

rafellerc avatar Nov 07 '18 16:11 rafellerc

Thank you! I will be waiting for your answer.

arbitularov avatar Nov 09 '18 05:11 arbitularov