Yahoo in conjunction flickr help push image recognition technology

The idea that image recognition would be another wave of the future was not uncommon until recently, November 28 news. But now, because of the size of the photos taken and stored and the need to find them quickly and easily, photographic technology pushes the image recognition technology to the forefront of development and the center of the stage. And as technology continues to evolve, image recognition technology is also adding other techniques to consolidating search operations and search results.

In recent days, Yahoo's Flickr and lab teams have developed a workable search algorithm using a multiple integration strategy that combines computer vision, geographic information and interpersonal interaction to get rid of metadata dependencies. This technology was developed to help optimize Flickr photo software as well as the user services for photo software, while also focusing on the current and future business needs of Yahoo Software and E-commerce services.

Yahoo Weather Software provides a good opportunity to experiment with this technology. The software collects metadata, geographic location, and people interacting with Flickr photos, thinking that the software chooses the most interesting pictures. "Weather software is a good example because it involves a lot of photos, and we can retrieve them in different locations, retrieve not only geographic coordinates, but also time and weather conditions, and then restore the images to the software," said David, senior research director at Yahoo Laboratories. And find out what kind of city you like. "Yahoo doesn't rely entirely on content from users, such as labels and photo names, and uses computer vision and deep learning techniques to identify objects on a picture," he said. The combination of computer vision and social vision enhances the quality of search results.

The social computing method produces about 6 million weather pictures. About 1.5 million photos remain on the weather software after correcting for inaccurate geographic location and time, low-resolution, and error content. "This approach strengthens artificial intelligence and adds to the human factor," said Jia Li, a senior research scientist at the Yahoo Lab. "Based on depth learning, the photo categorization algorithm analyzes whether a picture is a storm, a cloud or a snow map, day or night." ”

This is not the first time Flickr has publicly analyzed the details of its progress in image recognition technology. Another recent blog post also describes the team's attempts to improve image search and image-seeking techniques. Yahoo and Flickr combine social computing (social computing), geographic orientation, and computer vision to promote the development of image recognition technology. Yahoo also joined other companies to develop image recognition technology, sharing more than 100 million photos of its own collection to provide enough data for researchers to use.

Eventually, when the image recognition technology matures, it will help improve people's private life. Garrigues said, "Image recognition has not yet affected the lives of ordinary people." I have 10,000 photos on my phone, and I need to spend a lot of time sliding the screen, but I still can't find the photos I want. We believe that there is still a missing link in bringing this technology to people and helping people to improve their lives. "(Compilation: Cao Jian)

(Responsible editor: Lvguang)