VideoActionPredicto​r

VideoActionPredictor uses MATLAB to extract HOG features from videos, predict activities with an LSTM model, and send predictions to server

Overview:The VideoActionPredictor project uses MATLAB to process video files, extract features using Histogram of Oriented Gradients (HOG), and predict human activities using a Long Short-Term Memory (LSTM) neural network. The predicted activities are then sent to a FastAPI server.

Key Components:

1. Feature Extraction:

• The videoToFeatures function reads video files and extracts HOG features from selected frames. These features are essential inputs for the LSTM model.

1. Model Training:

• The script loads video data, extracts features, and trains an LSTM model on categorized activities such as 'fall', 'lie', 'nothing', 'sit', and 'walk'.
• Key parameters include the number of hidden units, dropout rate, batch size, and number of epochs.

1. Prediction and Classification:

• The check_directory function monitors a specified directory for new video files, processes each video to extract features, and uses the trained LSTM model to classify the activity.
• The predicted activity is then sent to a FastAPI server using HTTP POST requests.

1. FastAPI Server:

• The FastAPI server receives the predicted activities and can handle the results, such as storing them in a database or triggering specific actions.