27-30 Jul., 2015, Osaka



MIRU2015

The 18th Meeting on Image Recognition and Understanding

Tutorial

The day before the main conference, we will have the following tutorials.
Please note that the tutorials are presented in Japanese.

Date and Time
7/27 [Mon] 13:15-16:45
Venue
Main hall
Participation fee
Fee is included in the registration fee of the main conference.
Time Room A Room B
T1.
7/27 [Mon] 13:15-14:45
"Computational Photography: Recent Progress in Theory and Application"
Shinsaku Hiura (Hiroshima City University)
“Random forest and its applications in computer vision”
Hitoshi Habe (Kinki University)
T2.
7/27 [Mon] 15:15~16:45
“Recent Active 3D Scanning System and Techniques”
Hiroshi Kawasaki (Kagoshima University)
“Revisiting similarities and dissimilarities between patterns”
Toshikazu Wada (Wakayama University)


Tutorial 1A

“Computational Photography: Recent Progress in Theory and Application”
Shinsaku Hiura (Hiroshima City University)
Computational photography is a imaging technology which stands on the fusion of optical and computational processing, and it has around ten years history as an emerging research area attracting widespread attentions by many researchers. During the time, a lot of studies for deepening fundamental theory or aiming practicality have been made, and many products for industries and consumers have been commercialized. So, what have we found in this decade? Which direction should we go as our next step? In this tutorial, I firstly summarize the schematic view of computational photography technology and refer to the recent progress in theory and applications followed by the discussion on future directions.

Tutorial 1B

“Random forest and its applications in computer vision”
Hitoshi Habe (Kinki University)
Random forest is an ensemble learning method which consists of multiple decision trees as weak classifiers. It has been applied to various application areas because its structure is simple but effective and it can be used not only for classification but also regression and clustering. Human pose estimation by Shotton et al. is one of the epoch-making applications in computer vision community. In this tutorial, I will give an overview of random forest and its applications in computer vision.

Tutorial 2A

“Recent Active 3D Scanning System and Techniques”
Hiroshi Kawasaki (Kagoshima University)
Recently, various kind of 3D scanning systems are widely available, e.g., Kinect, Lidar, etc. Surprisingly, the number of papers using RGB-D sensor in CVPR is increasing and this year at least 19 papers include the related keyword in the title. Since basic techniques and algorithms of those systems are all different, accuracy, resolution, frame rate and other specification are also largely different each other. If users know the characteristics of those systems, it is preferable to find suitable system for their experiments. Further, novel techniques have been still intensively researched and proposed. In this tutorial, basic algorithms of current 3D scanning systems as well as practical implementation for the system will be introduced. In addition, future direction of active scanning systems and their applications will be also mentioned.

Tutorial 2B

“Revisiting similarities and dissimilarities between patterns”
Toshikazu Wada (Wakayama University)
Similarity or dissimilarity selection is essential for most pattern recognition and retrieval algorithms. However, most of these algorithms are opened to use any similarity or dissimilarity measures, and we often pay little attention to choose these measures. This tutorial shows possible measures defined between dense vectors, sparse vectors, and vector sets, and gives a selection guideline by showing their properties with some examples. Especially, we introduce “commonality” as a similarity measure and discuss the relationship with DNN.