CVPRF2026 EI, a new framework for multimodal medical image categorization.
ICASSP2025 DCP, a unified model that segments vessels in CFP, SLO, UWF, OCTA and FFA images.
ECCV2022 SuperRetina, end-to-end trained keypoint detection and description for retinal image matching.
Video-X retrieval
ECCV2022 LAFF, lightweight attentional feature fusion for text-to-video retrieval
ACMMM2022 Negation Learning, learning to answering video queries with negative cues.
TPAMI2021 Dual Encoding, dual encoding of video and text for text-video matching.
TMM2021 SEA, sentence encoder assembly for better query representation.
ACMMM2019 W2VV++, winning solution for the TRECVID 2018 ad-hoc video search task.
Data
Multimodal Large language models (MLLMs)
MICCAI2025 FunBench for evaluating fundus reading skills of MLLMs.
CVPR2025 PhD for evaluating visual hallucinations in MLLMs.
AI for health
MMM2026 OCTA60 for cross-modal fundus image registration.
JBHI2022 MMC-AMD for multi-modal categorization of age-related macular degeneration.
ICPR2020 Retinal-Lesions for retinal lesion segmentation and DR grading.
ACCV2008 Fundus10K for training and evaluating laser scar detection algorithms.
Cross-lingual / multi-lingual multimedia tasks
ACMMM2023 ChinaOpen, a new video dataset targeted at open-world multimodal learning, with raw data gathered from Bilibili.
TMM2019 COCO-CN, a bilingual image description dataset enriching MS-COCO with manually written Chinese sentences and tags, used for multiple tasks including image tagging, captioning and retrieval, all in a cross-lingual setting.
Video-X retrieval
ICCV2025 MGSV-EC, a large-scale dataset for Music Grounding by Short Video (MGSV).
TV2022 V3C1-PC, auto-generated video description dataset for pre-training text-video retrieval models.