Grupo de Tratamiento de Imágenes

 

 

News and Events 

AnimalMotionCLIP: Embedding Motion and Dense Frame Inputs in CLIP for Animal Behavior Analysis 

On June 18, Carlos R. del Blanco presented his innovative poster entitled "AnimalMotionCLIP: Embedding Motion and Dense Dense Frame Inputs in CLIP for Animal Behavior Analysis" at the IEEE/CVF Computer Vision and Pattern Recognition (CVPR) 2024 Conference, held at the Seattle Convention Center.

CVPR2024 V2

The project addresses animal behavior recognition through deep learning techniques using models such as CLIP. The main challenges are integrating motion information and designing an effective temporal model. AnimalMotionCLIP interleaves video frames and optical flow in the CLIP framework and compares several temporal modeling schemes: dense, semi-dense and sparse. Experiments on the Animal Kingdom dataset demonstrate that AnimalMotionCLIP outperforms current approaches.

CVPR is the foremost annual event in computer vision, offering excellent value for money for students, academics and industry researchers. Professor del Blanco's presentation has been very well received, highlighting its innovation and potential impact on animal behavior analysis.

 

CVPR2024 v1