7 Haziran 2014 Cumartesi

Animal Vision I

  It is not uncommon in computer vision literature to compare its results, methods with human vision. Well, it is so because of human vision is the 'best' vision we know most of the time thanks to its intelligence.  Yet, bio-inspired vision may benefit spending time on other spices. There are different solutions to different vision problems in nature. Therefore, I decided to learn a little bit more about how animals present solutions for their tasks.  
 Computer Vision community heard about  Mantis Shrimp perhaps numerous times.This amazing animal has 12  channels in its eyes, can see in range of 300 to 720 nm of wavelength(extended in both boundaries when compared to humans) ,has far more receptors than us, an exceptional vision system.

Mantis Shrimp

 However, I always wondered one thing - How such small brain can handle so much visual data. But apparently it perceives the surroundings in a different way than what we expected. In this work , Mantis Shrimps had troubles to distinguish between very similar colors, nearly performed as humans. So first of all, Manthis Shrimp scans object of interest with eye movements, number of visual receptor is not making it hard to process; vice versa, it makes it easier to process as more visual information is already gathered from sensory which is something our visual cortex has to deal with.

 Although it sounds a bit disappointing, imitating such system can be more helpful for vision society. Quite different spectral band cameras are available in the market and they are easy to reach; however, processing power is not yet there. In this case nature may lead us the by-pass this processing power constraint for some tasks.


24 Ocak 2014 Cuma

My Eigenfaces Presentation

Here you can find my presentation about famous paper "Eigenfaces for Recognition", I will soon upload my talk and notes.

 Link to slideshare

15 Kasım 2013 Cuma

Image Processing And Computer Vision Lectures

 UC Davis

 University of Central Florida CRCV; Lectures, thesis defenses, Conference Presentations

William Hoff's EGGN lecture series 510 and 512

Heidelberg University Image Analysis and Pattern Recognition

NPTEL Image Processing

Image and video processing: From Mars to Hollywood with a stop at the hospital 
Presented at Coursera by professor: Guillermo Sapiro of Duke university


Probabilistic Graphical Models from Stanford (Coursera)


From Manheim University

http://vision.in.tum.de/teaching/ws2013?redirect=1 (Many great CV lectures, check the teaching website)

11 Kasım 2013 Pazartesi