"If we knew what it was we were doing, it would not be called research, would it?"


Albert Einstein

Agent-based Learning to Utilize Local Data for Anomalous Activity Recognition (Funded by the U.S. Department of Homeland Security)


  • We aim to detect anomalous events or suspicious activities such as assault, explosion, and shooting in surveillance videos.​

  • We plan to improve the accuracy of decisions of human agents by reducing the manual work of monitoring of human agents.​

  • We focus to provide better visualization to locate anomalous event and act accordingly.​

House Detection from Aerial Images Using Faster RCNN

Demonstrate the effectiveness of Faster Region-based Convolutional Neural Network (Faster-RCNN) to detect buildings automatically from aerial images using Python programming language.

Key Notes:

  1. Implementation of faster RCNN algorithm.

  2. Data annotation using LabelMe.

  3. Bounding-box information extraction from XML file.

  4. Pre-trained ResNet50 for feature extraction.

Under Construction