Moodstocks SIFT on iPhone GPU

An iOS app featuring SIFT descriptors extraction with OpenGL ES 2.0

https://github.com/Moodstocks/sift-gpu-iphone

Automatic Summarization and Annotation of Videos with Lack of Metadata Information - Survey

Dear all,

I would like to ask you to participate in a survey. The goal of this survey is the collection of tags, which representatively describe the content of a video.

In the link: http://chatzichristofis.info/?page_id=677 each participant is called to watch five different videos and annotate them by recommending five representative keywords/ tags. These selected videos are among the most viewed videos according to YouTube statistics. With your assistance, the results of this survey will be used as a ground-truth set in order to evaluate our automatic video annotation method.

The estimated average time of the experiment is about 10 minutes. Thank you in advance for your help!

Seasons Greetings to everyone!!!!!

LuminAR bulb lights path to augmented reality

(Phys.org)—Are we moving closer to a computer age where "touchscreen" is in the room, but it is the counter, desktop, wall, our new digital work areas? Are we moving into a new form factor called Anywhere? Do we understand how locked up we are in on-screen prisons, and that options will come? The drive for options is strong at the MIT Media Lab, where its Fluid Interfaces Group has been working on some AR options, the "Augmented Product Counter" and the "LuminAR." The latter is a bulb that makes any surface a touchscreen. You can even use it to replace the bulb in a desk lamp with the MIT group's "bulb" to project images onto a surface. The LuminAR bulb is small enough to fit a standard light fixture.

The LuminAR team, Natan Linder, Pattie Maes and Rony Kubat, described what they have done as redefining the traditional incandescent bulb and desk lamp as a new category of "robotic, digital information devices." This will be one of the new looks in AR interfaces. The LuminAR lamp system looks similar to a conventional desk lamp, but its arm is a robotic arm with four degrees of freedom. The arm terminates in a lampshade with Edison socket. Each DOF has a motor, positional and torque sensors, motor control and power circuitry. The arm is designed to interface with the LuminAR bulb. The "bulb," which fits into a lightbulb socket, combines a Pico-projector, camera, and wireless computer and can make any surface interactive. The team uses the special spelling "LuminAR" to suggest its place in the group's other Augmented Reality initiatives.

Read more at: http://phys.org/news/2012-12-luminar-bulb-path-augmented-reality.html#jCp

www.lire-project.net

The site for the upcoming book "Visual Information Retrieval Using Java and LIRE" is on-line

http://www.lire-project.net/

Improving SURF Image Matching Using Supervised Learning

(Suggested Article)

Hatem Mousselly-sergieh [LIRIS] , Elod Egyed-zsigmond [LIRIS] , Mario Döller [FH Kufstein Tirol - UNIVERSITY OF APPLIED SCIENCES] , David Coquil [University of Passau] , Jean-Marie Pinon [LIRIS] , Harald Kosch [University of Passau]

Dans The 8th International Conference on Signal Image and Internet Systems (SITIS 2012), Naples, Italy.

Abstract

Keypoints-based image matching algorithms have proven very successful in recent years. However, their execution time makes them unsuitable for online applications. Indeed, identifying similar keypoints requires comparing a large number of high dimensional descriptor vectors. Previous work has shown that matching could be still accurately performed when only considering a few highly significant keypoints. In this paper, we investigate reducing the number of generated SURF features to speed up image matching while maintaining the matching recall at a high level. We propose a machine learning approach that uses a binary classifier to identify keypoints that are useful for the matching process. Furthermore, we compare the proposed approach to another method for keypoint pruning based on saliency maps. The two approaches are evaluated using ground truth datasets. The evaluation shows that the proposed classification-based approach outperforms the adversary in terms of the trade-off between the matching recall and the percentage of reduced keypoints. Additionally, the evaluation demonstrates the ability of the proposed approach of effectively reducing the matching runtime.