Speaker: Navneet Dalal, BotSquare, Inc.
Object detection in images is one of the core problems in computer vision with many applications such as in digital image analysis, video surveillance, smart cars and robotics to name a few. Two key components of existing state-of-the art object detectors are feature extraction and learning algorithms. In this presentation, we will explore the question of feature sets and how to use them for robust visual object recognition. We will cover the current state-of-the-art in feature sets such as Haar features, Histograms of Oriented Gradient descriptors, and the process of using these features for robust object detection. We will also cover key factors impacting the accuracy of the object detectors.
Speaker Bio
Navneet Dalal is founder & CEO of BotSquare, Inc. He has 11 years of experience in computer vision and has worked on many problems such as object detection, image search, image segmentation, panoramic mosaics from video sequences and image based rendering. Prior to starting BotSquare, he was Head of Computer Vision Research at Like.com (acquired by Google in 2010). Prior to that he finished his PhD in computer vision from INRIA, Grenoble, France. His PhD thesis research is the de-facto standard in computer vision for object detection in images & videos having received more than 2500+ citations in the last 5 years. It is currently being used by researchers the world over and is being taught as graduate curriculum in many prestigious universities like Stanford, Berkeley, and CMU. He is also author of 4 patents & 13 pending applications.
