OPPO demonstrates some of its potential new camera tech at CVPR 2020
Chinese smartphone manufacturer OPPO has demonstrated some of its potential new camera technologies at the Computer Vision and Pattern Recognition (CVPR) Conference 2020, which have earned the company two first-place and two-third place awards at the event. The technologies that won the awards include perceptual extreme super-resolution technology, visual localization, and human activity recognition in videos.
Perceptual Extreme Super-Resolution
During the event, OPPO’s Research team was challenged to solve the problem of perceptual extreme super-resolution for a single image with a 16x magnification factor. The aim of the challenge was to develop an AI model capable of producing high-resolution results with the best perceptual quality and similarity to the ground truth. For the challenge, OPPO’s team demonstrated the effectiveness of its RFB-ERSGAN neural network which delivered the following results:
OPPO’s neural network helped the company beat 280 other participants and secure the first place in the challenge. According to a press release on the matter, the neural network will be applied in a number of scenarios, including smartphones, where it will be able to help change low-resolution photos taken by the smartphone to high-resolution images with greater details. Other potential applications include restoring old or damaged photos, refining important medical images, or conducting a detailed analysis of satellite images.
Visual Localization for Handheld Devices
In the Visual Localization for Handheld Devices challenge, OPPO created a monocular visual localization pipeline that leverages semantic and depth cues to find the precise location of any given image. OPPO’s solution helped the company secure the first-place for outdoor visual localization and the third-place for indoor visual localization.
Potential use cases for this technology include augmented reality applications, intelligent robotics, or precise navigations. The company claims that an AI assistant equipped with a camera will be able to utilize the technology to analyze images and achieve accurate positioning, allowing it to guide users in the right direction.
Activity Detection in Extended Videos
OPPO also secured the third-place in the Activity Detection in Extended Videos challenge, which required companies to design automatic activity detection algorithms. OPPO’s solution is designed to effectively identify and reorganize the characters in each frame and recognize dozens of human activities in complex dynamic videos.
The technology is expected to play an important role in the field of human-computer interaction and sports video analysis in the future. Additionally, the technology may also be found in future smartphones that will allow them to automatically identify moving subjects and adjust the camera focus accordingly.