Challenges

Thumos logo We have obtained top ranking results in the Thumos 2014 and Thumos 2013 Action Recognition Challenge. The goal of the challenge is to evaluate large-scale action recognition in natural settings. The dataset used is the newly released UCF101 dataset, which is currently the largest action dataset both in terms of number of categories and clips, with more than 13000 clips drawn from 101 action classes.
TRECVID logo LEAR participated together with the AXES project to the TRECVID MED 2013 challenge, and finished in first position. The Multimedia Event Detection (MED) evaluation track is part of the TRECVID Evaluation. The goal of MED is to assemble core detection technologies into a system that can search multimedia recordings for user-defined events based on pre-computed metadata.

Modeling video dynamics and human actions (WP1)

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

Towards understanding action recognition

ICCV'13

Automatic learning of visual models (WP2)

MBH

Multi-fold MIL Training for Weakly Supervised Object Localization

CVPR'14 		Spatio-Temporal Object Detection Proposals

Spatio-Temporal Object Detection Proposals

ECCV'14
 

Joint learning from textual annotations and visual data (WP3)

Finding Actors and Actions in Movies

Finding Actors and Actions in Movies

ICCV'13 		Weakly Supervised Action Labeling in Videos Under Ordering Constraints

Weakly Supervised Action Labeling in Videos Under Ordering Constraints

ECCV'14

Active large-scale learning (WP4)

MBH

Good Practice in Large-Scale Learning for Image Classification

PAMI'14 		MBH

Transformation Pursuit for Image Classification

CVPR'14
MBH

Convolutional Kernel Networks

NIPS'14