Bertrand Luvison

The largest dataset for human-robot interaction anticipation reveals that existing models struggle with generalization in diverse real-world scenarios.

Controllable diffusion models can be trained twice as fast and achieve better image quality simply by directly supervising the clean target image during training.

Transformer-based object detectors are surprisingly resistant to adversarial attacks crafted for CNNs, but can be hardened further by training against a diverse "cocktail" of strong attacks.