ASHUTOSH MISHRA
RESEARCH WORK
Stereo correspondence algorithm
Developed a color based stereo correspondence algorithm that uses adaptive local support.
Like other color based stereo correspondence algorithms, it assumes constant color on the object surfaces, which significantly reduces the discontinuities in the depth map. Our algorithm is designed to achieve following goals: (a) speed: our algorithm is significantly faster than most of the other color based algorithms, (b) obtaining reliable depth at
object boundaries, (c) obtaining correct depth map for texture-less regions. The algorithm uses two different matching cost functions for high and low textured regions, thereby achieving trade-off between speed and accuracy. Like most other algorithms we use bidirectional matching to detect outliers.
The research paper describing the algorithm was selected for presentation in IEEE Symposium Series on Computational Intelligence, 2013, Singapore.
Hierarchical Temporal Memory
HTM is a machine learning algorithm developed by Numenta Inc. Hierarchical Temporal Memory (HTM) is a machine learning technology that aims to capture the structural and algorithmic properties of neocortex. HTM provides a theoretical framework for understanding the neocortex and its many capabilities. The algorithm can be used to learn and recall Spatial-Temporal Patterns. My work on HTMs is still premature.