On April 19, NASA announced that the Ingenuity helicopter successfully completed its historic first flight on Mars and landed safely, marking the beginning of a new era of “flight” for human deep space exploration and the use of AI. The motto of NASA’s Jet Propulsion Laboratory is Dare Mighty Things! and has become the slogan for mankind to step into the unknown and face challenges. What a fitting slogan to represent the field of AI and its myriad of possibilities!
Credit: NASA/JPL-Caltech/ASU/MSSS
On that note, and with the flight of the Ingenuity helicopter, three competitions held by maadaa.ai for CVPR 2021 are in full swing!
1. Short-video Face Parsing Challenge
Short video content is all the rage, used for social media, retail and so much more, AI technology has become a powerful tool in this booming field. Cool applications such as face recognition, virtual background, and cartoon images are all supported by AI technology. Among them, accurate recognition of facial information in videos is the most critical technology. maadaa.ai released 1400 short videos from our own dataset. The image data in the videos were collected, labeled, and divided into three parts: training set, valid set and test set. The challenge is now in progress! Please click “Read More” to obtain more details.
2. Human-centric Video Matting Challenge
With the recent pandemic, a majority of citizens around the world have been asked to work from home and international travel has been heavily restricted. Because of this, more and more people communicate through video conferencing apps. One of the most interesting options in these video conferencing applications is the virtual background, which can be changed at will. This technology uses AI to “segment out” the contour characters of the image. Based on this, maadaa.ai developed an open data set specifically for video conferencing scenarios to support innovative research on video conferencing-based background virtualization, semantic segmentation, etc. We used an online meeting APP to collect data in real scenes such as offices, cafes and parks. The challenge is in progress, please click “Read More” to obtain more details.
3. High-resolution Human Parsing Challenge
Data plays a very important role in the development of AI technology. An algorithm model design requires a large amount of labeled data for training. The rise of large-scale annotated data sets has raised many possibilities in the field of computer vision. However, in some fields, such as the health care and insurance industries, the high cost of algorithm research and development caused by data shortage has become an unavoidable problem in the development of AI. Through this competition maadaa.ai hopes to bring together researchers from related industries to discuss how to develop the application of computer vision technology in the context of limited or imperfect data and to explore related emerging technologies. The challenge is in progress, please click “Read More” to obtain more details.
