One paper, MagicBokeh, was presented as a CVPR 2026 Oral paper and listed as a Best Paper Finalist by public paper indexes.
Bo Li
Principal Researcher, Camera Research Department, vivo · Head of vivo BlueImage Lab
I lead research at vivo BlueImage Lab, where we work on mobile imaging, computational photography, content generation, visual agents, and foundation models. My current work connects academic research in vision and generative models with camera, album, and on-device imaging systems.
Before vivo, I was a researcher at Tencent Youtu Lab. I received my B.Sc. and Ph.D. degrees from Nanjing University. My research has appeared in CVPR, NeurIPS, ICLR, ICML, ICCV, ECCV, IJCV, TPAMI, AAAI, and ACM Multimedia, spanning AI imaging, dense prediction, camouflaged object detection, federated learning, adversarial robustness, diffusion models, and multimodal vision.
I also serve the research community as a Senior Area Chair for NeurIPS and Area Chair for ICLR. Public profiles: Google Scholar, OpenReview, DBLP, X.
news
Four vivo BlueImage Lab papers were accepted to ICLR 2026, covering reference-based diffusion super-resolution, video bokeh, temporal video generation, and live-photo restoration.
BlueImage Lab had seven papers across NeurIPS, ICCV, ICML, ICLR, and AAAI, including two NeurIPS papers and two ICCV papers.
Representative 2024 work included three CVPR papers, three ECCV papers, two ACM Multimedia papers, plus IJCV and NeurIPS work.
selected publications
all publications
Federated Learning with Label Distribution Skew via Logits Calibration
ICML 2022 · federated learning under label-distribution skew
DENSE: Data-Free One-Shot Federated Learning
NeurIPS 2022 · data-free federated learning
Revisiting Single Image Reflection Removal in the Wild
CVPR 2024 · computational photography and restoration
Towards Photorealistic and Efficient Bokeh Rendering via Diffusion Framework
CVPR 2026 Oral · vivo BlueImage Lab · MagicBokeh
vivo BlueImage Lab
The lab's public research agenda covers mobile imaging, computational photography, content generation, agents, and large models. Recent open projects include MagicBokeh, AdaRefSR, Any-to-Bokeh, SDMatte, PPC, and SmartPhotoCrafter.