Deep learning video exercise classification

Sir, it is possible by using Matlab's deep learning model. We train through our desired video and later test it with another video. Please guide me on how it is possible.

[Christopher Stapels]

Can you describe more what you wish to do? I'd start by looking at the deep learning in MATLAB page. There is also a page on deep learning with video. You have ThingSpeak tagged, but you cannot upload videos to ThingSpeak at present. Did you mean to involve ThingSpeak in your workflow?

[Utpal Das]

I want to check if the exercise is done correctly or not from my uploaded video. By applying some algorithm

[Christopher Stapels]

Are you attempting to upload a video to ThingSpeak, or are you trying to use MATLAB in ThingSpeak to process your video?

Im trying to clarify if your question involves our IoT platform.

[Utpal Das]

I am trying in MATLAB

[Christopher Stapels]

Ok thanks, I rremoved the ThingSpeak tag to help other users. Good luck with your project.

[Utpal Das]

Thank you for telling me about Thingspeak. Is there any possibilities to do with Thingspeak

Answer 1

[Rahul]

There are variety of application you can do with deep learning. As per your question, the outcome that you want to achieve is not clear. I am assuming you want to classify by giving the video input as training. In supervised classification, I assume you have labeled video that you can use for training. You can extract the frames of a recorded video and then based on the extracted images, you can train your CNN model. Below is the code for how to extract frames from a recorded video file. The variable "all_frames" contains alll the frames present in a video. Use these frames to train your CNN model. The related documentation links are given below:

Create object to read video files - MATLAB - MathWorks India

Deep Learning in MATLAB - MATLAB & Simulink - MathWorks India

Classify Videos Using Deep Learning - MATLAB & Simulink - MathWorks India

Once the training is complete, you can use the same code to extract images of a test video and predict your outcome using a trained model. Hope this helps.

P.S.: Use MATLAB R2016a or later.

vidObj = VideoReader('xylophone.mp4');

if contains(vidObj.VideoFormat, 'RGB')
    % This is created if the video is of format RGB.
    all_frames = zeros( vidObj.Height, vidObj.Width, 3, vidObj.NumFrames, 'uint8' ); 
else
    % This is created if the video is of format grayscale.
    all_frames = zeros( vidObj.Height, vidObj.Width, vidObj.NumFrames, 'uint8' ); 
end

ii = 1; 
while( hasFrame( vidObj ) )
    frame = readFrame( vidObj );
    if length( size( all_frames ) ) == 4
        all_frames(:, :, :, ii) = frame;
    else
        all_frames(:, :, ii) = frame;
    end
    ii = ii + 1;
end

imshow( all_frames( :, :, :, 1 ) ) % displaying 1st frame