src

Dataset

• https://www.dropbox.com/sh/75v5ehq4cdg5g5g/AABvnJSwZI7zXb8_myBA0CLHa?dl=0
• https://www.kaggle.com/datasets/mission-ai/crimeucfdataset/

Installation

pip install -r requirements.txt

Observations

https://github.com/Hyouteki/BTP/blob/main/ucfc/OBSERVATIONS.md

Quick Start

python extfet.py
python makmodred.py

Basic file overview

• extfet: Extracts frames from the dataset and saving it to features.npy file. Reduces the overhead of extracting frames in each test.
• model_conv_lstm: Trains the ConvLSTM based model on the dataset with reduced classes.
• model_lrcn: Trains the LRCN based model on the dataset with reduced classes.
• OBSERVATIONS: Observations and details rearding each test of the model.
• models/: Trained models.
• ind_model: Contains observations regarding models trained to detect only a signle class.
• trainfiles/old/: Depricated models.

Dependencies

• Tensorflow
• Keras
• opencv
• Termcolor

References

• https://github.com/WaqasSultani/AnomalyDetectionCVPR2018
• https://bleedaiacademy.com/human-activity-recognition-using-tensorflow-cnn-lstm/
• https://youtu.be/QmtSkq3DYko?si=6VzZc_NH5glCPi0m
• https://learnopencv.com/introduction-to-video-classification-and-human-activity-recognition/