Venue: KD 101
Date: 23rd Nov 2015
Time: 15:30 hrs

Title

-------

Expanded Parts Model for Human Analysis and Nonlinear Models for Classification and Embeddings


Abstract

--------------					

						

In the first part of the talk I will introduce our Expanded Parts Model (EPM) for recognizing human attributes (e.g. young, short hair, wearing suit) and actions (e.g. running, jumping) in still images. An EPM is a collection of part templates which are learnt discriminatively to explain specific scale-space regions in the images (in human centric coordinates). EPM uses only a subset of the parts to score an image and scores the image sparsely in space, i.e. it ignores redundant and random background in an image. To learn our model, we propose an algorithm which automatically mines parts and learns corresponding discriminative templates together with their respective locations from a large number of candidate parts. 

In the second part of the talk, I will introduce our novel approach for learning nonlinear support vector machine (SVM) corresponding to commonly used kernels in computer vision, namely (i) Histogram Intersection, (ii) χ2, (ii) Radial Basis Function (RBF) and (iv) RBF with χ2 distance, without using the kernel trick. The proposed classifier incorporates non-linearity while maintaining O(D) testing complexity (for D-dimensional space), compared to O(D × Nsv ) (for Nsv number of support vectors) when using the kernel trick. We also promote the idea that such efficient nonlinear classifier, combined with simple image encodings, is beneficial for image classification. In an extension to this work, we show how nonlinear embeddings can be learnt to compress image signatures for efficient semantic category based image 
 retrieva
 l
.



References (preprints available from www.grvsharma.com/publications.html)

----------------------------
Part I

Expanded Parts Model for Semantic Description of Humans in Still Images

G. Sharma, F. Jurie, C. Schmid

arXiv:1509.04186

Saarbrücken, Germany, Sep 2015

(Extended version of CVPR 2013)
Part II

Learning Nonlinear SVM in Input Space for Image Classification 

G. Sharma, F. Jurie, P. Perez

Technical report, hal-00977304

Rennes, France, 2014 

(Extended version of BMVC 2013) 

Scalable Nonlinear Embeddings for Semantic Category-based Image Retrieval

G. Sharma, B. Schiele

International Conference on Computer Vision (ICCV)

Santiago, Chile, Dec 2015 (to appear)

Biography

--------------

Gaurav Sharma is currently at the Max Planck Institute for Informatics, Germany. He holds an Integrated M.Tech. (5 years programme) in Mathematics and Computing from the Indian Institute of Technology Delhi (IIT Delhi) and a PhD in Applied Computer Science from INRIA (LEAR team) and the University of Caen (CNRS GREYC Lab), France. He was a Senior Engineer at the Technology Planning Group of Samsung Delhi R&D before starting his PhD and was a Researcher at the Exploratory Research Group at Technicolor Rennes R&I after obtaining his PhD. His primary research interest lies in Computer Vision and Machine Learning for tasks such as image classification, object recognition and facial analysis.