45 Denys Rozumnyi Computer Vision News Computer Vision News precise analysis of the video, we are interested in reconstructing the object in 3D with its full 6 DoF motion with sub-frame precision (3D reconstruction). The proposed pipeline for motion-blurred object processing is useful for a variety of applications like sports analysis, detecting fast-moving obstacles in front of a vehicle for driver alert and autonomous driving, and general video quality enhancement or compression. This work could eventually make motion-blurred objects processing accessible to everyone using regular cameras rather than expensive high-speed cameras and push the capabilities of current high-speed cameras to a new level, e.g. by capturing gun shots. As an actual use case, we refer to a popular drone footage that claims to contain an UFO. This footage has been viewed by millions of users. A user on a social media platform posted a high-impact post, saying about this analysis: “some madman did the math... like ALL THE MATH... and now I think this sighting is truly incredible”. However, another independent user accessed Denys’s method and showed that the UFO footage, in fact, contained a diving maneuver of a gyrfalcon, one of the world's fastest birds, reaching speeds of up to 209 km/h. This use case has gained significant media attention (even Forbes talked about this and called Denys an UFO hunter). Some users doubted the output of the method, claiming that the method was presumably trained on a dataset that contained such a bird. However, it was trained on a synthetic ShapeNet dataset without any living creatures. The suggestion to train on sci-fi movies would be considered. UKRAINE CORNER
RkJQdWJsaXNoZXIy NTc3NzU=