Ongoing Project

Find connections & communities of friends from personal photo data which makes easy to:
- Find particular photos of friends and communities
- See who knows who and who should be invited (e.g. when you plan a party)
- Hide particular connections & communities from the visualization results (e.g. while using this technique as GUI of SNS)
However it might be redundant and complicated to visualize the person who belongs to the multiple communities. Our challenge is to visualize effectively using actual photos.

Past Project

Face Image Processing
Various kinds of techniques for recognition of facial expressions have been recently presented. Those techniques realize better human and computer interaction. In most of the techniques, computers recognize facial changes by tracking face areas, facial parts and features. Those techniques will realize that computers understand human’s facial expression, psychology, gestures, and so on. Some researches have attempted to recognize facial changes by detecting of particular facial feature points and tracking them.
However, those techniques have problems that elaboration and robustness for human’s motions are not enough. When we consider the human in the natural state, we also have to consider sudden occlusions and various human’s motions. Robustness of feature tracking is crucial to recognize natural facial expressions and movements.
Our technique first detects particular feature points by template matching, and then tracks the points in moving images by optical flow calculation. At the same time, the technique estimates the reliabilities of the detected feature points. If a feature point seems not to be reliable, the technique executes template matching in the neighbor region or approximates the position of the feature point. By those steps, the technique is capable both robust and quick manners.

(Using a video data of NRC-IIT Facial Video Database)


Past Project

Visualization of Impression Evaluation Results of Women's Makeup
Recently, many studies have examined the effect of facial makeup on the impression of the face. However, these experimental results are really complicated and difficult to clearly understand because various factors affect impressions of makeup faces. We present a technique to visualize correlation between makeup faces and impressions in this poster. Our technique helps users to understand the results by using GUI which extended CAT (Clustered Album Thumbnails) an image browser to visualize a large amount of pictures.

(Left) Examples of images of a woman with/without makeup. The upper image is without eyeliners, while the lower image is with long eyeliners. (Right) Visualization of impression evaluation result with various eyeliners.

Visualization of impression evaluation result with various eye shadows.

Our implementation for selection of particular images.

Media Art Works

Worked on some media art products in the past. Will be updated soon.