Qinlei Huang

RankVR achieves a significant leap in Composed Image Retrieval by effectively managing noisy data, outperforming state-of-the-art methods in challenging scenarios.

ReTrack tackles the inherent bias towards reference videos in composed video retrieval by cleverly calibrating directional biases in composed features, leading to state-of-the-art performance.

HABIT achieves superior image retrieval performance by simulating human habit formation, effectively tackling the Noise Triplet Correspondence problem that plagues traditional methods.

INTENT achieves robust image retrieval by effectively mitigating both cross-modal and modality-inherent noise, outperforming traditional methods in the presence of annotation errors.