kegangwangccnu/physbench

Simple, fast, and fair evaluation of remote physiological sensing models

physbench's Issues

inference using other models

hi,
thanks for you work.
i was wondering if there is a way to get inference using physformer, i tried but it s kinda complicated, can you help , thanks

the model weight of Physformer

Hello, first of all, thank you very much for your open-source code!
It seems that you are involved in the Physformer model here. I met the question that the rppg shape is really strange，from the Physformer‘s github code：ZitongYu/PhysFormer#21

Do you have the model weight of Physformer trained by yourself and how can I use it?

Thanks for sharing!!!

Data Usage Agreement

Can you upload a template file of the Data Usage Agreement for RLAP? Thanks.

A question of EfficientPhys

Hello, first of all, thank you very much for your open-source code! Hello! It seems that you are involved in the EfficientPhys module here. Do you have the source code for EfficientPhys?

报错：模型预期有0层，但是预训练权重有9层

运行命令：python inference.py --video 1424.avi --model efficientphys --show-wave发生了两个报错：

1. 模型中有一个名为 'ep' 的层，这个层的 build() 方法被调用了，但是这个层并没有实现 build() 方法。这可能会导致一些问题，因为这个层会被标记为已经构建（built），尽管它实际上并没有被构建。
2. 模型预期有0层，但是预训练权重有9层

记录一下报错记录：

PS D:\桌面\PhysBench\PhysBench> python inference.py --video 1424.avi --model efficientphys --show-wave
2024-04-06 15:10:53.108186: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation
orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0.
2024-04-06 15:10:53.789180: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation
orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0.
2024-04-06 15:10:55.480582: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
D:\Anaconda\envs\physbench\lib\site-packages\keras\src\layers\layer.py:357: UserWarning: build() was called on layer 'ep', however the layer does not have a build() method implemented and it looks like it has unbuilt state. This will cause the layer to be marked as built, despite not being actually built, which may cause failures down the line. Make sure to implement a proper build() method.
warnings.warn(
Traceback (most recent call last):
File "D:\桌面\PhysBench\PhysBench\inference.py", line 53, in
ep.load_weights('./weights/EfficientPhys_UBFC.h5')
File "D:\Anaconda\envs\physbench\lib\site-packages\keras\src\utils\traceback_utils.py", line 122, in error_handler
raise e.with_traceback(filtered_tb) from None
File "D:\Anaconda\envs\physbench\lib\site-packages\keras\src\legacy\saving\legacy_h5_format.py", line 357, in load_weights_from_hdf5_group
raise ValueError(
ValueError: Layer count mismatch when loading weights from file. Model expected 0 layers, found 9 saved layers.

is there pre-trained model -> efficientPhys

hi!

is there efficientPhys pre-trained model?

Missing timestamp content in utils.py?

In dump_datatape there should be a store for the timestamp.

如何对VIPL数据进行处理呢？

您好，我看在你的项目中有一个PhysFormer_vipl.ipynb文件，里面是直接使用训练好的权重。想请教一下如何使用vipl进行训练呢，我看在数据处理中并没有对VIPL进行处理。感谢您的回复。

A question about conference.py

Hello, first of all, thank you very much for your open-source code! I am a student and I need to refer to your code recently. I have a question about your code.
model = EP() model.build(input_shape=(None, 450, 8, 8, 3)) model.load_weights(args.weights)
I would like to use EfficientPhys here. How can I modify the code？

Inquire saved values via inference.py

Hello, First of all, I would like to thank you for sharing your valuable knowledge and results.
I am a student to use the model implemented in your repository.
In the code implemented on inference.py , there are lines 68, 69

predict = np.nanmean(_, axis=0)
bvp = np.concatenate([bvp, predict])

Can you tell me what the bvp value obtained through this code means? When I read read.me , it's called the wave value, but I don't know because I'm not familiar with this domain.

Which code should I refer to if I want to extract and save my heart rate?
I'd appreciate your help.

请问训练时是否进行了人脸和关键点检测？比较困惑

您好，在学习您这个框架过程中，发现在公开数据集的预处理中，在load_dataset函数中进行了人脸以及关键点检测。但最后生成.h5文件中并没有写入boxes和landmark数据，这两个数据都放在tmp文件夹中的缓存文件中。
后续在推理过程中（inference.py）使用了人脸以及关键点检测。
因此比较困惑，想知道是否在训练过程中 进行了人脸以及关键点检测，是否使用了tmp中的缓存文件。
感想您的解答，谢谢