zhang-pengyu / hmft Goto Github PK

Thanks for your great work! When will the code be available? Thanks again.😊

Hello, thank you very much for the dataset and benchmark code you provided. I have a question to ask you. How do you deal with the frames where the objects disappear after the objects disappear in the image during the training phase? I would be very grateful if you could get back to me!

Hello! Thank you very much for open source code, so that I can learn more.
I would like to ask you more details about the training, what form does your dataset exist in? For example, is it in that folder, or is there any format requirement for the folder? Thank you very much!

First of all, thanks for the good research.

What part of the code is the multi-step training process described in the paper and in what order do i run the code files?

Can you provide a more detailed description of the learning process in your readme file?

Where can we find the scripts to evaluate the predicted results by the pre-trained model ? When the running the prediction scripts (for example run_RGBT210.py), it only generates the predictions of the model on that dataset. No performance metrics are generated.

In the line,
https://github.com/zhang-pengyu/HMFT/blob/215cb846fc5da59e4e7810ac92bdc085840b0270/mfDiMP/ltr/models/fusion_model/feature_fusion.py#L31C7-L31C7
,
Operations should not be performed on batch_size dim,
I think it may be torch stack
Can you answer my doubts?

您好，请问图片下方链接提供的追踪结果和表格中的指标一致吗，我用VTUAV数据集的toolkit在链接中的结果上运行，得到的指标和表格给出的62.7，75.8差别比较大

您好，请问您们会开源训练代码嘛，谢谢！

请问有没有在LasHeR数据集上进行过测试？

The following links to download demos seems to be broken:

https://drive.google.com/uc?id=1LeYyYxPA5XBF4m73T1p9qXsmGKVV7Fmm
https://drive.google.com/uc?id=1yrtTYKcUfNsk-jxjXmY54FIxtBS3Fz9H
https://drive.google.com/uc?id=1uiqFuG3RnWcyu-OaRzKIQhrCLXKKM7_z
https://drive.google.com/uc?id=1fL4Yo4xBSndeE_ShAqrraiHAL7QvrEAi

Error message:

     $ gdown https://drive.google.com/uc?id=1LeYyYxPA5XBF4m73T1p9qXsmGKVV7Fmm -O mfDiMP/demo/bus1.zip

     Access denied with the following error:
   
     Cannot retrieve the public link of the file. You may need to change the permission to 'Anyone with the link', or have had many accesses.

     You may still be able to access the file from the browser:

     https://drive.google.com/uc?id=1LeYyYxPA5XBF4m73T1p9qXsmGKVV7Fmm

您好，我对您的工作非常感兴趣，有以下几点疑惑想咨询您：

1. 论文中提到的多步训练的三个训练阶段的训练epoch数分别为多少？或者在您给出的代码中在哪里体现了多步训练吗？
2. 最后一步训练ADF模块和IOU预测模块时，论文中提到要固定所有backbone并微调classifiers，那DFF模块需要微调吗？

Hello,
When I tried to run demo.py I had the folowwing error:

Traceback (most recent call last):
  File "demo.py", line 20, in <module>
    tracker.parameters.features.features[0].dis_net_path = 'OptimTracker_ep'+str(45).zfill(4)+'.pth.tar'
AttributeError: 'Tracker' object has no attribute 'parameters'

I believe that the conda environment is created and all the mecessary packages have installed succusfully. Any advice would be useful for me.
Thanks in advance

While trying to run the demo.py, we get the following error:

        Cuda Error : RuntimeError: CUDNN_STATUS_EXECUTION_FAILED

Looks like there is an mismatch in the cuDNN version to be installed. Can you provide the cuDNN version which was used to train the model 'OptimTracker_ep0045.pth.tar' ?

您好，想问一下作者，VTUAV数据集采用稀疏标注，即每十帧给一个bbox的标注信息，但是如果sampler采样到未标注帧的frame_ids，您是怎样处理的呢？在您的项目里，没有找到该部分对应的代码段。可以提示一下您获取未标注帧目标框bbox信息，代码所在的位置吗，非常感谢！！！