zitongyu / physformer Goto Github PK

Hello, I want to test on my own dataset, but the document named "VIPL_fold1_test1.txt" is missing. What is the specific function of this document? and how do I prepare this documentation for my dataset? Look forward to your reply.

Hi, would it be possible for you to please share your code for pre-processing of the video clips (i.e. MTCNN implementation, downsampling the frames, etc.). Thank you

Hello, thank you for provide the PhysFormer code. I have a question about Physformer, so I'm asking 3 questions.

1. How were the frames for each video made?
   Currently I'm using ffmpeg to cut video into frames with fps. But I'm wondering if there is another way that you used.

2. About VIPL_fold1_train.txt
   I'm wondering why the fps of certain video in VIPL_fold1_train.txt is lower than 25(such as 16). According to the VIPL-HR dataset, it says the fps of videos are 25, 30 but it seems like the fps in VIPL_fold1_train.txt are varying. I wonder how to manipulate raw fps and ppg values.

3. About VIPL_fold1_train.txt 2
   I found out that the ppg wave in VIPL-HR is different to VIPL_fold1_train.txt. I'm assuming that some butter bandingpass filter is used, still I want to know how the ppg wave is processed.

I am very interested in your work

Is this method able to measure Heart Rate (HR) in real-time?

Hi, when attempting to train Physformer out of the box it seems to fail because it is missing a path to input_data. The default is set to "/scratch/project_2003204/" which is not found in the repo. I was wondering if this folder would be made available?

Thank you.

Inference_HRevalutaion.m file only contains the MAE could you also give the code for SD, RMSE, and another doubt I have is whether we are only considering average HR of each video for error calculation?(I did not find anywhere the use of values present in VIPL_fold1_test1 after the average_HR)

Could you either elaborate or share your code on how frequency based HRV features were calculated?

Sorry for disturbing, I just found out that the VIPL dataset has published data of testset in fold1.

I'm having trouble with the TDC module,the paper explains that it uses the TDC module proposed by AUTOHR.However, I don't think CDC_T code implements the formula for TDC in AUTOHR.Why use out_normal-06*out_diff, and out_diff is to first calculate the combination of the weights of the t0 fragment and the t2 fragment in the convolution kernel, and then use the 3D convolution to volume, I am very confused, can you explain,thank you

Hi,

Do you have a timeline for your code release?

Thanks.

Thanks for sharing the code, may i ask the google drive pretrained model weight is training based on 160 frames or 220 frames?
Because when I test inference_OneSample_VIPL_PhysFormer.py, I found in Loadtemporal_data_test.py the variable clip_frames=220 while in Loadtemporal_data.py clip_frames=160.
On the other hand, i print the each start id and end id of clip, I found every clip exsists overlap of some images, but dont have in training code Loadtemporal_data.py. May I ask the reason of seting the overlaps in Loadtemporal_data_test.py?

following is the printed image id of Loadtemporal_data_test.py

start image id:  61
end image id:  281
start image id:  221
end image id:  441
start image id:  381
end image id:  601
start image id:  541
end image id:  761

In this equation, i am confused what is k. I think YHR is heart rate value for a given subject. but its not clear to me what is k.

I read the paper Age Estimation Using Expectation of Label Distribution Learning

here lk is true age and y is probability distribution function

Dear author, our team is doing an academic research on the extraction of physiological features of the characters in the video, in the process of collecting data, I saw your paper PhysFormer: Facial Video-based Physiological Measurement with Temporal Difference Transformer, which is very helpful for our research, we are very interested in your research on HR, RF, HRV on OBF data set, and would be grateful if you could provide us with the relevant source code and weight files.

As the title says, could I know how to procure OBF dataset since it's a perfect dataset for research on Physiological Measurement.

When dealing with data and label alignment, there are some problems. I can't get the bvp label in VIPL_fold1_train.txt accurately, how do you align when processing the data? If possible, is it possible to provide the complete txt file. thank you.

Hi, thank you for the code, I have a question.
Why is there no HR1 in the Matlab file generated after testing the data?

Hi,

Thank you for the awesome work!

Is there any plan for the release of PhysFormer++?

Thank you!
Kind regards,
Dylan

hello, I run inference is error, how to fix?

我们训练的时候 需要把VIPL数据集都变成图片吗？下载下来的 数据集是视频格式的 但是仔细产看代码 里面都是对图片进行处理。