The code for EpicFlow (CVPR'15) is available here.

Motion Boundaries

The detection code for "Learning to Detection Motion Boundaries" (CVPR'15) is available here.

Convolutional Kernel Networks

The code for our NIPS'14 paper on convolutional kernel networks is available on the project web page.

Kernel Temporal Segmentation

The code for "Category-specific video summarization" (ECCV'14) is available here.


The code for our CVPR'14 paper on transformation selection is available on its project page .

SPAMS v2.5

The open-source sparse estimation toolbox SPAMS v2.5 has been released. It contains the code for the CVPR'14 paper on archetypal analysis, and the ICML'13 and NIPS'13 papers on large-scale optimization.

Human Pose Estimation in Videos

The code for our CVPR'14 paper Mixing Body-Part Sequences for Human Pose Estimation, is available on the project page.

Segmentation Driven Object Detection with Fisher Vectors

The code for the ICCV'13 paper Segmentation Driven Object Detection with Fisher Vectors by R. Gokberk Cinbis et al. is available here: MaskFishDet.


The code for our Bioinformatics paper on fast RNA-seq isoform deconvolution is available on the project web page and also as a Bioconductor R package.


Code for DeepFlow, an optical flow estimation algorithm based on DeepMatching (see below) and published in our ICCV'13 paper, is available here.


Code for DeepMatching, a dense and robust matching algorithm mentioned in our ICCV'13 paper, is available here.

Improved Trajectories Video Description

The code for improved trajectories in our ICCV'13 paper - can be found on the website of Heng Wang.

Stochastic Gradient Descent for multiclass linear classification

The Stochastic Gradient Descent package mentioned in Perronnin et al. CVPR 2012 is available here.

INRIA's Fisher implementation

Here is an implementation of Florent Perronnin's Fisher aggregation method for local descriptors.

Dense Trajectories Video Description

The code for dense trajectories in our CVPR'11 paper - can be found on the website of Heng Wang.

Metric Learning for verification and Multiple Instance Learning

Matthieu Guillaumin's LDML and MildML on his code page.

Lear's GIST implementation

This library is a C reimplementation of A. Torralba's GIST descriptor. It was used in our CIVR 2009 paper.


The weighted nearest-neighbor model for multiple binary or multiclass classification that we successfully used it for image auto-annotation in our ICCV'09 paper is available here.

Spatio-Temporal Descriptor for Action Recognition (HOG3D)

The software for computing the spatio-temporal descriptor using gradient orientations and regular polyhedrons (or platonic solids) - as described in our BMVC'08 paper - can be found on the website of Alexander Kläser.

Trecvid example generator

This program generates example videos for the Trecvid copy detection task.

Groups of adjacent contour segments

Software for detecting Vitto's local shape features (see bottom of page).

Face Recognition

Xiaoyang Tan and Bill Triggs' MATLAB face normalization and descriptor code, as described in our AMFG'07 paper.


YORG, developed within the AceMedia project helps organizing photo collections.

Object Detection and Localization Toolkit

You may download the Navneet Dalal's object detection and localization toolkit from here.

Binaries and test images for Krystian Mikolajczyk's scaled and affine interest point detectors and various types of local image descriptors.

Gyuri Dorko's version of these detectors.

Texture Based Image Classification

Frank Moosmann's FRIMCLA Randomized forest based image classifier, as described in our ECCV'06 workshop and NIPS'06 papers.

Interest Point Detectors & Test Sequences

A joint comparison of affine covariant regions by LEAR, Oxford, Leuven and Prague (the test setup and the binaries are available)