Abstract: Text-to-video generation is getting attention and the generated videos can be used in many applications. However, it is uncertain whether existing deep learning techniques work well for generated videos. In this paper, we compose a study of how generated videos can be retargeted by deep learning models with the ratio of the main object preserved and looked for ways to improve the quality of the generated and retargeted video frames. Throughout the experiment, we discover the errors of video retargeting on the generated videos in the processes of segmentation, inpainting, and relocating.
External IDs:dblp:conf/icufn/KimKP23
Loading