Research Directions

Thoth's research on learning based approaches for visual scene interpretation can be divided into the following areas. For more information see our publications.

Action recognition in videos

Dense trajectories and motion boundary descriptors for action recognition

Dense trajectories and motion boundary descriptors for action recognition

IJCV'13
Mixing Body-Part Sequences for Human Pose Estimation

Mixing Body-Part Sequences for Human Pose Estimation

CVPR'14
 
MBH

DeepFlow: Large displacement optical flow with deep matching

ICCV'13
MBH

Temporal Localization of Actions with Actoms

PAMI'13

Weakly-supervised visual modeling

MBH

Multi-fold MIL Training for Weakly Supervised Object Localization

CVPR'14
Finding Actors and Actions in Movies

Finding Actors and Actions in Movies

ICCV'13
 
MBH

Label-Embedding for Attribute-Based Classification

CVPR'13
Spatio-Temporal Object Detection Proposals

Spatio-Temporal Object Detection Proposals

ECCV'14

Large-scale classification

MBH

Large-scale image classification with trace-norm regularization

CVPR'12
MBH

Good Practice in Large-Scale Learning for Image Classification

PAMI'14
MBH

Optimization with First-Order Surrogate Functions

ICML'13
MBH

Convolutional Kernel Networks

NIPS'14

Large-scale retrieval

MBH

Aggregating local image descriptors into compact codes

PAMI'12
MBH

Product Quantization for Nearest Neighbor Search

PAMI'11
 
MBH

Event retrieval in large video collections with circulant temporal encoding

CVPR'13
MBH

Combining attributes and Fisher vectors for efficient image retrieval

CVPR'11