Pattern Recognition and Machine Learning Lab

Projects

Semantic image and video analysis

  • Object detection and scene classification
  • Object co-occurrence modeling for anomaly detection
  • Video summarization based on semantic co-occurrence modeling
  • Incremental modeling of scene statistics
  • Publications: MTA (2013), PR (2012), IJCNN'12
Wildlife

Event classification / anomaly detection

  • Optical flow extraction
  • Manifold modeling of motion patterns using spatio-temporal locality preserving projections and Laplacian eigenmaps
  • Online, incremental manifold learning
  • Publications: AI13, IVCNZ'12
STLPP

Distributed system analysis, modeling and optimization

  • Smart sensing
  • Energy conservation and coverage preservation
  • Predictive data aggregation using light-weight sensor implementations
  • Optimized solutions for routing and streaming in multi-hop wireless networks
  • Recent publications: MLSDA'13, ISSNIP13, HPCC'12, ISSNIP'11.
Sensing

Video segmentation and object extraction

  • Adaptive algorithms based on pixel modeling using mixture of Gaussians, Codebook, and Growing Neural Gas
  • Adaptive tuning of parameters
  • Superpixel-based object extraction
  • Details see MVA, AI'13, IVCNZ'12/11, ICONIP'11, ACCV'12 workshop (2nd best paper prize)
BG