• Publications
  • Influence
Reconfigurable models for scene recognition
TLDR
We propose a new latent variable model for scene recognition. Expand
  • 167
  • 14
  • PDF
Improving bag-of-features action recognition with non-local cues
TLDR
We propose to disambiguate local space-time features and to improve action recognition by integrating additional nonlocal cues with BoF representations in a kernel-combination framework. Expand
  • 130
  • 11
  • PDF
Automatic Discovery and Optimization of Parts for Image Classification
TLDR
We unify the two stages of learning part-based models and learn the image classifiers and a set of shared parts jointly. Expand
  • 54
  • 8
  • PDF
N4: a precise and highly sensitive promoter predictor using neural network fed by nearest neighbors.
Promoters, the genomic regions proximal to the transcriptional start sites (TSSs) play pivotal roles in determining the rate of transcription initiation by serving as direct docking platforms for theExpand
  • 15
  • 2
  • PDF
Spotlight the Negatives: A Generalized Discriminative Latent Model
TLDR
We introduce generalized discriminative latent variable models (GLVM) which use both “positive” and negative latent variables. Expand
  • 12
  • 2
  • PDF
Generalized Majorization-Minimization
TLDR
We introduce Generalized Majorization-Minimization (G-MM), an iterative bound optimization framework for non-convex optimization, and show empirically that they consistently outperform their MM counterparts in optimizing non-Convex objectives. Expand
  • 4
  • 1
  • PDF
Reading Street Signs using a Generic Structured Object Detection and Signature Recognition Approach
TLDR
We address the applied problem of detecting and recognizing street name plates in urban images by a generic approach to structural object detection and recognition. Expand
  • 6
  • PDF
Image Classification with Reconfigurable Spatial Structures
We propose a new latent variable model for scene recognition. Our approach represents a scene as a collection of region models (“parts”) arranged in a reconfigurable pattern. We partition an imageExpand
  • 1
  • PDF
Modeling Image Context Using Object Centered Grid
TLDR
We propose an alternative approach to retrieval of contextual information, by customizing the location of the grid based on salient objects in the image. Expand
  • 4
  • PDF
Reducing Ambiguity of Local Descriptors for Visual Recognition
One of the most commonly used methods for representing visual information is to study properties of image regions at local neighborhood of pixels. Visual appearance, by its nature, is subject toExpand