KTH Computer Vision Group

Group's Research Results

Object Class Recognition

Mixture component learning for object recognition (ECCV '12)

Strong Supervision for Part Based Models (ECCV '12)

Human Pose Estimation

Automatic Multiple View Human Pose Estimation (BMVC '13).

Note that the final pose is estimated independently for each frame.
With manual labour and without calibration information! (SCIA '11).

With no calibration and no manual effort (BMVC '12).

Facial Landmark Localisation

Find parts and multiple regressors to find landmarks (BMVC '10).

Fun stuff! - Robot Sketches a Face

The robot arm makes a sketch of this face.

(Details in Radu Pana's masters degree project )

Thanks to the robotics group for their help!

Current Projects

VINST/ Visual Diaries

Recent developments in proccessing units, memory and storage devices and image processing and computer vision algorithms makes it possible to process large amount of data automatically: using a normal PC, it is possible to store more than a thousand hours of videos in storage devices and process millions of images to find similar images automatically in a second. Also, the mass production and cheap prices of cameras has made it possible to collect large amounts of images and videos without any significant cost. Utilizing such developments, we decided to collect, store and process videos similar to visual input of some subjects and study to what end we can use such a data to help people with limited memories. In other words, our long term goal is to answer the question of What is worth remembering in a data set of every day life videos?
Project page

Freeview Immersive Networked Experience

The Free-viewpoint Immersive Networked Experience (FINE) project will be focused on researching and developing a novel end-to-end architecture for the creation and delivery of a new form of live media content. FINE will introduce the concept of live free-viewpoint content which will provide rich and compelling immersive experiences by allowing remote viewers to place a virtual camera in a real live-action scene and move it freely in space and time, heightening their sense of presence and reality.
Project page

Completed Projects

Capturing and visualizing large scale human action

The goal of this project is to build a 3D reconstruction of the motion of all the players during a professional fooball game. This has numerous potential applications especially for the visualization of the game for spectators and its transmission for broadcastors. And also it would be of great interest to the games industry to obtain high quality motion capture from real games.
Project page