Computer Vision News - February 2018

Image compression is a key procedure for any image today. Without it, each 12-megapixel image would require 36 megabytes of storage space , rendering it extremely tedious and resource- consuming to upload, download, exchange and store any digital photography. Compression techniques have given satisfactory results during decades, with standards like JPEG and others. The organizers of the challenge believe that breakthrough neural networks techniques can grant additional advancement to modern image compression standards. Hence, they are calling the machine learning community to join this effort. A training set of 1,633 uncompressed images (see them at the bottom of this page) has been recently released and the test set will be released on February 15. Competing team are requested to submit the compressed versions of the test set by April 22 (extended deadline); rankings will be established based on both PSNR and human rating by experts. The overall winner will be decided by an ad hoc panel which will take into account also the runtime performance . Provided dataset is actually double: set P (“professional”) and set M (“mobile”). The two sets are meant to be representative for images commonly used in the wild. Participants can choose to train neural networks or any other methods on provided datasets and/or additional data, such as ImageNet and the Open Images Dataset . 26 Computer Vision News Challenge: CLIC Challenge Every month, Computer Vision News reviews a challenge related to our field. If you do not take part in challenges, but are interested to know the new methods proposed by the scientific community to solve them, this section is for you. This month we have chosen to review the CLIC - Learned Image Compression workshop and challenge, organized around CVPR 2018 which will be held later this year in Salt Lake City, Utah. The website of the challenge, with all its related resources, is here . The challenge is co-sponsored by Google, ETH Zurich and Twitter. Workshop And Challenge On Learned Image Compression