kalpitthakkar/pb-gcn

Code for the BMVC paper (http://bmvc2018.org/contents/papers/1003.pdf)

pb-gcn's Issues

about path

Could you give me a detailed explanation of the path，
For example,work_dir， data_path ，missing_txt and split_dir

ValueError

Hello, thank you for your work.
When I tried to train the dataset while running the program, the following error occurred:

Traceback (most recent call last):
File "D:/UserData/yanhan/RE-ID CO/pbgcn/pb-gcn-master/pb-gcn-master/run.py", line 366, in
part=part
File "D:\UserData\yanhan\RE-ID CO\pbgcn\pb-gcn-master\pb-gcn-master\utils\nturgbd\prepare_ntu.py", line 74, in gendata
os.path.join(data_path, s), max_body=max_body, num_joint=num_joint)
File "D:\UserData\yanhan\RE-ID CO\pbgcn\pb-gcn-master\pb-gcn-master\utils\nturgbd\read_skeleton.py", line 43, in read_xyz
seq_info = read_skeleton(file)
File "D:\UserData\yanhan\RE-ID CO\pbgcn\pb-gcn-master\pb-gcn-master\utils\nturgbd\read_skeleton.py", line 7, in read_skeleton
skeleton_sequence['numFrame'] = int(f.readline())
ValueError: invalid literal for int() with base 10: '\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x0

I suspect there is a problem with the dataset. Is the file structure of the dataset as follows?
--nturgb+d_skeletons
--S001C001P001R001A001.skeleton
--S001C001P001R001A002.skeleton
--S001C001P001R001A003.skeleton
......

I hope my problem can be solved，thanks.

Signals?

Hello, I hope to kown the definition and function of signals computed in the rel_angles.py and orient_disps.py file, respectively? And what is the performance of these additional signals?

MemoryError

Hello, there is a MemoryError in the runtime. What is the cause of this situation? I have 32 GB of memory and run time memory utilization is almost 100%.

How much GPU memory needed?

I tried to run training with batch size 1 or 2, but it shows "Cannot allocate memory" error after occupying some GPU global memory around 5~700MB. I have 2 GPUs-GTX1080.

Thank you~!

Channel setting in config file

Hi, great work. Why did you set your input channel as number 15 (for example in nturgbd_xsub_train.yaml, model_args.channel)? Isn't it 3 channels for (x, y, z)？ Or you just copy the 3 channels for 5 times as input? Thanks.

question about Hdm05

I got a question for you, How can I execute the HMD05 dataset?

Questions about HDM05 dataset

Hi, thanks for your great contribution!
Could you give me some details about how to preprocess the HDM05 dataset or offer processed .mat files of HDM05 dataset? In fact, when I was going to preprocess the HDM05 dataset, I couldn't tell which data (the full takes or the cuts) did you download and whch file (.c3d or .amc) did you use to generate .mat file?

what is for "reference point"?

Hello,
I found "references=[4,8,12,16]" is used to get relative loc/angles.
each point corresponds to face, left hand, right hand, left foot in NTU-RGB+D dataset figure

I'm trying to apply pb-gcn on COCO style skeleton(17 point).

should I use "[0, 7, 4, 13]" for coco 17 keypoints to match NTURGBD style skeleton?

Plus, I have wanted to know how to get the weighted sum of each features of body parts using eq(16) in your paper.

below code lists in "st_gconv_resnet.py" are for that?

        # model
        x = self.head(x)

        for layer in self.layers:
            x = layer(x)

        # V pooling
        x = F.avg_pool2d(x, kernel_size=(1, V))

        # M pooling
        c = x.size(1)
        t = x.size(2)
        x = x.view(N, M, c, t).mean(dim=1).view(N, c, t)

        # T pooling
        x = F.avg_pool1d(x, kernel_size=x.size()[2])

        # C fcn
        x = self.fcn(x)
        x = F.avg_pool1d(x, x.size()[2:])
        x = x.view(N, self.num_class)

Thank you~!