goutamyg / smat Goto Github PK

Hello, thanks for great work!
I run the code from [OSTrack/tracking/profile_model.py] to measure the model parameters and MACs .(https://github.com/botaoye/OSTrack)
Howerer, overall params is 739.205K, which is different from 3.8M in the paper.
The experiments.yaml and lib/config/* are the same as Main branch(8b99155).
Could you please tell me how to measure the model size?

When I run the "video_demo.py" file, an error message appears: KeyError: 'target_response, can you help me check it?

Can you provide me with a pre-trained model named "mobilevitv2-1.0.pt"?Thank you very much!

Hello, I downloaded the corresponding pre-trained model "mobilevitv2-1.0.pt" from [ml-cvnets] and configured the relevant parameters. However, during the training process, there are the following issues:
"SerWarning: An output with one or more elements was resized since it had shape [8388608], which does not match the required output shape [1, 128, 256, 256]."“
This warning covers almost every layer. How should I correct it? Thank you for your support and work.

Hello, how did you draw a heatmap? The effect I drew using the Gradcam method is much worse than the one in your paper. Would you like to make this part of the code public?

Hii @goutamyg
While training, On the first epoch only, I am getting the same type of error as mentioned here. Also after commenting the lines told by you, i am still getting the same error, can you help me with it. Thank you
RuntimeError: Given groups=1, weight of size [256, 128, 1, 1], expected input[128, 3, 128, 128] to have 128 channels, but got 3 channels instead

When I tested the results using the "vis_result. py" file, I encountered an error message: AttributeError: 'VisResults' object has no attribute 'visdom'. Can you help me check this issue?

When I train to epoch 10, I will encounter an error message, which is “RuntimeError: Given groups=1, weight of size [32, 3, 3, 3], expected input[1, 128, 32, 32] to have 3 channels, but got 128 channels instead”. Can you help me solve this problem? Thank you.

Hey, your work looks impressive!
I wanted to fine-tune SMAT on a custom dataset. However, changing all the configurations and files seems a little complex. Is there any chance that training on a custom dataset feature is also added in the codebase? Thanks!

Hi, how can I find the confidence score or how can I calculate it?
Thank you.