Human Activity Recognition Implementing CNN-LST

Human activity recognition is an important field in computer vision with numerous applications such as surveillance, activities, sports analysis, and event recognition. Over the years, Machine Learning has contributed powerful techniques such as Convolutional Neural Networks (CNN), Graph Neural Networks, and Long Short Term Memory (LSTM). In this report, we implemented a classification model that combines the strengths of CNN and LSTM networks to recognize human movement in video frames. This model consists of a comprised CNN that extracts visual features from video frames and an LSTM that learns the temporal relationship between the frames by using the UCF dataset, which allows for comparison with similar datasets in other studies. The simulation results have shown that the CNN-LSTM model is capable of achieving a high performance over passing some models that only use individual or hybrid techniques. This reflects the efficiency of the model for detecting human activity in video frames, as well as its potential for practical uses.