AI erases objects in footage in real time, making filming more 'leisurely'

Serhii Postupaiev, a member of the research team, said that Yolov8 works by dividing the video frame into a grid of cells, then predicting, classifying and removing unwanted objects that appear in each of these cells. .

The name Yolov8, short for "You Only Look Once", was developed by the research team based on the idea of ​​thoroughly handling footage accidentally captured by the cameraman or inappropriate objects. This technology will help with live broadcast events.

The research team used a huge data set, up to thousands of gigabytes, including videos recorded from many different devices to train AI. The tool used in the training process is Video Inpainting, a deep learning technology that works similarly to the Spot Healing Brush (a tool to remove objects on photos) of Photoshop software.

Postupaiev said that thanks to AI, technicians can easily remove redundant details in footage, making live broadcasts more professional and smoother, without being interrupted by unexpected incidents. want. This technology also makes the video editing process simpler.

Yolov8 will be applied first in the field of football, focusing on eliminating distracting details during live coverage. In the future, the research team will expand the application of this technology to many other fields.

Yolov8 is considered the first complete AI model capable of processing live video, although similar software has been developed before.

Samuel Daniel

Update 16 July 2024