1. ICNS Home
  2. Projects
  3. Current PhD Projects
  4. Hardware Implementation of Unsupervised Feature Extraction Algorithms

Hardware Implementation of Unsupervised Feature Extraction Algorithms

Research Student

Philip Chennakudy Jose

Research Project Details

Feature detection is a fundamental problem in the study of computer vision task. The main issue here is, how the features are extracted and used for building reasonable information. Over the years conventional image or video processing techniques have been used for extracting the features. In addition, those techniques are based on frame based processing producing reasonable and accurate results. However, those techniques have failed in using temporal content of the features.

Neuromorphic engineering is a paradigm shift happened in the field of engineering developing novel bio-inspired sensors which operates in a very different principle compared to the conventional frame based cameras. Instead of producing frames, those sensors produce events with high temporal precision. Moreover, the events records the pixel intensity changes. The conventional feature extraction algorithms are not designed to extract useful features at event-based processing domain. To address this drawback many algorithms are recorded in the literature. 'FEAST' is an unsupervised algorithm developed to tackle such problems in event-based processing domain. My research focuses on developing novel digital VLSI architectures for this algorithms which will be capable of accelerating this algorithms in real time.

About ICNS

Partnership

Research

Projects

Resources

Master's Program

Opportunities

Events

News

Contact