Before, we were more exposed to self-timer software that beautified photos. With the development of the times, retouching only for static photos can no longer meet the needs of video platforms, so the live broadcast "slimming" function was born, which allows you to slim your waist into a lightning bolt when you broadcast live, and then you can get breast enhancement! 36D is not a dream! Not a dream! Jinhua ham doesn't exist, it's all long-legged sisters.
Although everyone is using it, few people know that these seemingly simple functions are actually applications of artificial intelligence.
Under the background of Yan value economy, the contradiction between the traditional photography service mode and the needs of multiple users has become increasingly prominent. Based on the foresight and profound understanding of the development law of the industry, deep black technology has keenly captured the pain points of this industry, and started with artificial intelligence to upgrade the photography industry in an all-round way. Hunan Shenhei Technology Co., Ltd. is a research and development institution focusing on high-quality image processing and artificial intelligence application technology. Based on computer deep learning algorithm and image algorithm, it provides high-quality and fast artificial intelligence retouching service for the retouching needs of photography institutions such as portrait photography, news live photography, product photography and party building photography. On May 5th, 20 18, Deep Black Technology just won the third prize of Wuhan Station of 20 18AWS Smart Star Entrepreneurship Competition.
"Artificial intelligence retouching robot" is a high-speed automatic retouching cloud platform specially developed by Deep Black Technology for professional portrait photography institutions: wedding photography, costume photography, portrait photography, Han Feng wedding dress, parent-child photography, news photos, meetings, group building activities, competitions and other live photos. It can realize automatic selection of scrap pieces, automatic color matching, repair of worn tops, beautification and shaping, layout design and so on. Various photos taken by photographers and mobile phones are uploaded to the artificial intelligence retouching robot platform, and retouching can be easily and instantly completed without manual intervention. How to make the machine work like a retoucher? Deep black technology has the following four core technologies:
1, human key point detection
Accurately detect 182 key points of the whole body, including five senses, face, trunk, limbs and fingers. It can stably capture various postures of the whole body in the state of photos and videos. It can deal with complex situations such as clothing occlusion, trunk self-occlusion, multi-person occlusion and environmental occlusion. Multi-person gestures can be detected at the same time. Application: automatic whole body plastic surgery, automatic skin repair, automatic defect repair and automatic composition.
2, pixel-level image segmentation
Accurately identify nearly 100 kinds of objects, optimized for professional photography industry, and can identify objects of various scales in professional photography with huge size. Application: automatic whole body plastic surgery, automatic defect repair, automatic skin repair, automatic color matching and automatic background change.
3. Deeply understand the pictures
Understanding the content of a picture can give it meaning. Including the center of gravity, visual focus, character emotions, behavioral intentions, theme types, and picture atmosphere. Understand the picture order and context. Only by correctly understanding pictures can we beautify them to the extreme. Application: Intelligent screening, automatic color matching and automatic background changing.
4. Aesthetic engine
Imitate the perceptual ability of human brain, and comprehensively consider the information of expression, composition, main color, color harmony, saturation, contrast, dynamic range, subject clarity and other dimensions. Always perceive the effect of retouching and adjust the retouching techniques in real time, and stop at perfection. Application: automatic color matching, automatic typesetting and automatic typesetting.
In this era of popular retouching, retouchers have become a scarce resource, which is not a small cost. According to industry sources, taking the wedding industry as an example, in the wedding photography scene, the cost of retouching can account for 6%-8% of all costs.
Artificial intelligence retouching is a big weapon to solve the above problems. Artificial intelligence retouching is actually to realize retouching through skills, and then solve the problems of time cost and labor cost in the later stage of photography. At present, the processing speed of deep black technology is about 0.3 seconds, and the server can process more than100000 pictures a day. Many merchants said that if there is mature artificial intelligence retouching, they are willing to pay for it. They even said that when artificial intelligence retouching is mature, it will be a subversion of the whole retouching industry, and many people will be laid off.
In the future, the development of artificial intelligence retouching will gradually develop from relatively standardized products to more complex photo processing, such as from standardized photo processing in public relations activities to photo processing in fashion blockbusters, different photo studio retouching styles and even the photographer's personal style. If this level is reached, artificial intelligence retouching will obviously usher in the ultimate "transformation"!
In Bian Xiao's view, artificial intelligence retouching for B-end can achieve fine processing of images when its products are really mature, and perhaps even the C-end market can win.