ROBUST FINE-GRAINED OBJECT CLASSIFICATION IN VIDEOS DOMAIN
Fine-grained object classification task distinguishes between classes in a sub-category of objects, for instance, the particular species of bird or dog. It uses information from still images and ignores the information that can be exploited from the input video. Videos are more challenging than still images because they have many noisy frames that contain unfavorable poses, motion blur, problematic lighting, and occlusions. Over the years, several approaches and datasets have been proposed for fine-grained categorization in still images, but the enormous potential of video data to improve vehicle ...
(For more, see "View full record.")