Normal Vs. Robbery

This dataset is a refined subset of the UCF Crime Dataset, specifically focusing on Robbery and Normal activity classes. Each class contains 150 high-quality preprocessed videos, making it an ideal dataset for machine learning, deep learning, and computer vision applications such as anomaly detection and human activity recognition.

Key Features of the Dataset

✔ Preprocessed & Standardized – Each video is resized to 600 frames (~20 seconds) while retaining complete contextual information.
✔ Balanced Dataset – Includes 150 Robbery videos and 150 Normal videos, ensuring equal representation.
✔ Optimized for AI & ML Models – Ready-to-use for TensorFlow, PyTorch, OpenCV, and deep learning frameworks.
✔ Simplifies Anomaly Detection – The original UCF Crime videos were long and untrimmed, making them difficult to work with. This dataset eliminates that complexity.
✔ Efficient Sampling Techniques – Uses linear and random sampling to preserve essential details while reducing video length.
✔ Real-World Applications – Ideal for security surveillance, human behavior analysis, and crime detection systems.
✔ Compatible with Various AI Tasks – Supports classification, prediction, and real-time monitoring systems.

 Applications of This Dataset

🔹 Abnormal Human Behavior Recognition & Prediction
🔹 Robbery Recognition & Prevention Systems
🔹 Anomaly Detection in Surveillance Videos
🔹 Real-Time Crime Monitoring & AI-driven Security
🔹 Deep Learning Research in Computer Vision

This dataset is sourced from Kaggle.