Generating Bird’s Eye View from Egocentric RGB Videos

Figure 11

Comparison of the results from image-to-image and video-to-video translation methods. (a) The SSIM values of each generated frame with its corresponding ground-truth frame. The SSIM values in (a) for the image-to-image method do not seem to follow any trend, whereas for the video-to-video translation method, the quality of the image seems to degrade a little as more frames are generated. (b) The SSIM values of each generated frame with its previous generated frame. In (b), the consecutive frames from image-to-image translation show little similarity, whereas the consecutive frames from video-to-video translation show high similarity and hence consistency.