If i am not wrong then in your code you are taking only first 16 frames of each video and ignore remaining....Is there any reason for this?

Hi

I have made a python script to extract frames for multiple extenstions, but i dont delete the files once i convert the videos, but that can be done with an additional line of code.

frames =5 
video_folder = '/media/ryan/shakira/3d-cnn-action-recognition/some_folder'
for file in os.listdir(video_folder):
   if file.endswith('.webm'):
       os.mkdir(os.path.join(video_folder,file.split('.')[0]))
       folder_name = file.split('.')[0]
       os.system('ffmpeg -t 4 -i {} -vf fps={} {}/%05d.jpg -t 10'.format(os.path.join(video_folder,file), frames,os.path.join(video_folder,folder_name )))
   elif file.endswith('.mp4'):
       os.mkdir(os.path.join(video_folder,file.split('.')[0]))
       folder_name = file.split('.')[0]
       os.system('ffmpeg -t 4 -i {} -vf fps={} {}/%05d.jpg -t 10'.format(os.path.join(video_folder,file), frames,os.path.join(video_folder,folder_name )))
   elif file.endswith('.blob'):
       os.mkdir(os.path.join(video_folder,file.split('.')[0]))
       folder_name = file.split('.')[0]
       os.system('ffmpeg -i {} -vf fps={} {}/%05d.jpg'.format(os.path.join(video_folder,file), frames,os.path.join(video_folder,folder_name )))
   else:
print('*********************************************************', file)

Hi

Thanks a lot for sharing your work!

Could you please tell me what is the difference between the main 3dcnn.py and the one in 3dcnnv2.

Thanks

Hello,
I tried working on v2 of your 3dcnn but it gives me this error:

    main()
  File "3dcnn2_M.py", line 187, in main
    args.output, args.color, args.skip)
  File "3dcnn2_M.py", line 120, in loaddata
    return np.array(X).transpose((0, 2, 3, 4, 1)), labels
ValueError: axes don't match array

Any solution to that?

thanks

amax@amax:/media/amax/Masters/Hairen/ActionRecognition/3d-cnn-action-recognition/3d-cnn-action-recognition-master$ python3 3dcnn.py --batch 32 --epoch 50 --videos dataset/ --nclass 10 --output 3dcnnresult/ --color True --skip False --depth 10
/home/amax/anaconda3/lib/python3.6/site-packages/h5py/init.py:34: FutureWarning: Conversion of the second argument of issubdtype from float to np.floating is deprecated. In future, it will be treated as np.float64 == np.dtype(float).type.
from ._conv import register_converters as _register_converters
Using TensorFlow backend.
/home/amax/anaconda3/lib/python3.6/importlib/_bootstrap.py:219: RuntimeWarning: compiletime version 3.5 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.6
return f(*args, **kwds)
0%| | 0/1 [00:00<?, ?it/s]UCF-10
Traceback (most recent call last):
File "3dcnn.py", line 237, in
main()
File "3dcnn.py", line 145, in main
args.output, args.color, args.skip)
File "3dcnn.py", line 101, in loaddata
X.append(vid3d.video3d(v_file_path, color=color, skip=skip))
File "/media/amax/Masters/Hairen/ActionRecognition/3d-cnn-action-recognition/3d-cnn-action-recognition-master/videoto3d.py", line 29, in video3d
frame = cv2.resize(frame, (self.height, self.width))
cv2.error: OpenCV(3.4.3) /io/opencv/modules/imgproc/src/resize.cpp:4044: error: (-215:Assertion failed) !ssize.empty() in function 'resize'

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 11.82it/s]

Can you please tell how many numbers of GPU are required for the training purpose??
Or Any specification of GPU??

I want to use the same work that you have done but on another dataset (the MSR Daily 3d dataset)
the problem is I did'nt know how to extract the classe name and save them into a file text because de dataset is organised as this :

-video
-MSRDailyAct3D_pack1
-a01_s01_e01_rgb.avi
-a01_s01_e02_rgb.avi ........
.
.
-MSRDailyAct3D_pack8
so for every pack two actions are performed exemple pack1 (eat and drink)
I really need some help if you could sugget what should I do to generate class names for my dataset ?

python 3dcnn.py --batch 4 --epoch 100 --videos UT Interaction/ --nclass 6 --output 3dcnnresult/ --color True --skip False
File "", line 1
python 3dcnn.py --batch 4 --epoch 100 --videos UT Interaction/ --nclass 6 --output 3dcnnresult/ --color True --skip False
^
SyntaxError: invalid syntax

what i'm doing wrong please help

Hello @dipakkr @venkateshgupta5

Can i test the model using webcam.I need suggestions.

I tried to run 3dCNN-v2/3dcnn.py.
python3 3dcnn.py --batch 4 --epoch 100 --videos /home/user/3d-cnn-action-recognition/data2/ --nclass 101 --output 3dcnnresult/ --color False --skip False
In the middle of loading data, it suddenly Killed.
(Only "Killed" showed)
I used UCF101 dataset.
Why this happens?

When I input my dataset and I print X in line 102 I get arrays of zeros. it seems that what VideoCapture returns in line 101 is arrays of zero. Why this happens? Does the code able to read my videos or not when arrays are zero?How can I see the input video frames to make sure? I am running 3dcnn.py

I needed to test the model n some sample videos. Could please tell me how to do that