Nicolas Pugeault

This work generates synthetic plankton imagery conditioned on taxonomy: a CLIP encoder is adapted on a large plankton corpus with a ranked contrastive objective extended to deep, ragged taxonomies, then frozen to condition a parameter-efficient diffusion transformer.

Vision foundation models can effectively track animal poses with minimal labeled data, achieving superior accuracy and cross-species robustness.

MMPM reveals that accurately modeling pedestrian intentions can drastically enhance trajectory prediction, outperforming traditional methods by addressing multimodal behaviors directly.