How to video classification using feature extraction

Hello. I want to classify videos using several feature extraction.

I used HOG(histogram of oriented gradients), optical flow as feature extraction.

https://www.mathworks.com/help/vision/ref/extracthogfeatures.html

However, the accuracy of classification is not good. just 60% accuracy.

Could you give some ideas to improve accuracy?

This code is using HOG.

clear all
close all

%// read the video:
list = dir('*.avi')
% loop through the filenames in the list
for k = 1:length(list)
    reader = VideoReader(list(k).name);
    vid = {};
    while hasFrame(reader)
        vid{end+1} = readFrame(reader);
    end
    for i=1:25
        fIdx(i) = i; %// do it for frame 1 ~ 60
        frameGray{i} = rgb2gray(vid{fIdx(i)});
        [featureVector{i},hogVisualization{i}] = extractHOGFeatures(frameGray{i});
        
    end
end
    X = cell2mat(featureVector');

[Kenta]

related to your question, a video classification with CNN and LSTM would be very effective.

Pls find the example code below.

https://jp.mathworks.com/matlabcentral/fileexchange/74402-video-classification-using-lstm-lstm

[Ramesh M]

But in this above document with CNN and LSTM have error on activation function.

Kindly giev me solution

Answer 1

[Shishir Singhal]

For video classification, you can use CNN for extracting spatial features. CNN is capable to extract deep features that HOG and other handcrafted feature extraction techniques might not be albe to. Use LSTM for capturing temporal features beacause you also need to have some sequential information between frames in a video.

You can read about CNN and LSTM in links here :

https://www.mathworks.com/solutions/deep-learning/convolutional-neural-network.html

https://www.mathworks.com/help/deeplearning/ug/long-short-term-memory-networks.html