Neural network learned to estimate the beauty of photographs

SAN FRANCISCO - April 6, 2017 - PRLog -- Today new technology for estimating photographs based on their aesthetic value was launched. Everypixel trained the neural networks to see the beauty of photos the same way as you do. The beta version (https://everypixel.com/aesthetics) is available you can check how it works by uploading your images.

The algorithm estimates visual quality of an image and applies aesthetic score to every file. "We asked designers, editors and experienced stock photographers to help us generate a training dataset to provide neural network with an example to follow," says Everypixel CEO Dmitry Shironosov. "Neural networks have processed 956,794 positive and negative patterns and our algorithm has learned to see the beauty of images in the same way as you do."

In the near future this data will take part in the overall mix of ranking factors of each image to move it closer to the first page and weed out lower quality images from the search results. It also regularly renews search results to meet recent trends and minimise the large photo base microstock agencies accumulate. This new technology eliminates the trashy stock images that designers and photo editors are forced to scroll through in order to reach a quality image.

"Despite the enormous amount of images available online, the collections of majority stock photo sites are often overlapped. It is hard work to find a rare or fashioned stock photo. However, there are niche image sites with original images, which few people know of," Shironosov continues. "EveryPixel solves these problems by allowing customers to browse quality collections of different stock photo websites that they would never have found."

Shironosov has been in the Stock Photography Industry since 2006, launching a number of production studios with high-quality photos and video footages for top leading stock image agencies. However, after noting that a lot of stock images are of low quality, the launch of Everypixel creates an aesthetically filtered search engine to aggregate images from 50 of the world's leading microstock agencies.

The demand for still images is growing. In fact the large stock photography industry is estimated to have a gross revenue between 2.5 to 3 billion dollars. Shutterstock, one of the market leaders, generated 425.1 million in revenue in 2015. However, there is a huge oversupply. Many image searchers spend hours researching for the right image for their project as they battle with too many useless choices. Buzzfeed (https://www.buzzfeed.com/daves4/unexplainable-stock-photo...) assembled a small handful of these appalling examples, highlighting the tiresome excess needed to be searched through to reach anything of quality.

Major image search engines such as google or bing do not supply tools for a search across licensed content and specific stock image search engines such as librestock only offers free content.