Research Interest

I am interested in different problems in Computer Vision ranging from object detection, image semantic segmentation to understanding complex human activities. Below is a short summary of my current research.

• Human Activity Detection - While recognizing simple human actions has seen considerable success, understanding complex human activities that involve multiple objects/agents with concurrent and interleaved sequences still remain very challenging. My current research focuses on temporal and spatio-temporal detection of human activities in untrimmed and long video sequences (e.g., surveillance camera feeds) using Deep Learning.

• Object Detection - Though object detection has matured with the success of Deep Learning, detection of multiple objects under moderate to heavy occlusion is still a notorious problem. Moreover, generalizing over different scales of objects especially smaller objects is difficult to match. I am motivated to design occlusion-aware and scale-invariant deep object detectors that can deliver real-time performance.