Computer Vision & Multimodal Learning

Feihong Shen沈飞鸿

I focus on VLLMs Computer Vision and AIGC, exploring how machines can understand, retrieve, and generate visual content.

View resume GitHub Google Scholar

fhshen2022@gmail.com Hangzhou, China

Feihong Shen at Yamdrok Lake — Yamdrok Lake, Tibet

01 / About

Research with a focus on useful visual intelligence.

I am a Multimodal LLM Algorithm Engineer at Douyin E-commerce, ByteDance. I received my M.S. in Software Engineering from Southeast University in 2025, where I worked in the PALM Lab under Prof. Hao Chen. I received my B.Sc. in Artificial Intelligence from Southeast University in 2022.

My work spans VLLMs, Multimodal Content Understanding, AIGC, Self-Supervised Learning and RGB-D perception.

Multimodal LLMs
AIGC
Computer Vision
Self-supervised Learning

02 / Experience

Engineering & research

Jun 2025 - Present

Multimodal LLM Algorithm Engineer

Douyin E-commerce, ByteDance · Hangzhou, China

Upgrade the video product-matching pipeline with product-image recognition in video, video and product multimodal-LLM embedding retrieval, and video-to-product identical-item reranking.
Build product-to-video RAG retrieval for high-performing ad creative replication, and train post-hoc advertising performance estimation models.

Jul 2024 - Oct 2024

Computer Vision Algorithms Intern

Tongyi Lab, Alibaba Group · Hangzhou, China

Developed camera-motion-controllable image-to-video generation methods based on diffusion models.

03 / Education

Southeast University

2022 - 2025

M.S. in Software Engineering

School of Computer Science and Engineering

First-Class Si-Tech Scholarship, 2024
Second-Class Academic Scholarship, 2022 & 2023
Second Prize, China Post-Graduate Mathematical Contest in Modeling, 2022
Teaching Assistant for Computer Vision, Fall 2022

2018 - 2022

B.Sc. in Artificial Intelligence

School of Computer Science and Engineering

GPA: 3.8 / 4.0 · Rank: 24 / 86
Merit Student, School of Computer Science

04 / Selected work

Publications

Research on multimodal visual understanding and content-aware generation.

IEEE TIP 2024CCF A

Disentangled Cross-modal Transformer for RGB-D Salient Object Detection and Beyond

Hao Chen, Feihong Shen, Ding Ding, Yongjian Deng, Chao Li

A cross-modal transformer that disentangles shared and complementary features, reaching strong results across eight RGB-D benchmarks.

Paper Code

NeurIPS 2024CCF A

Prune and Repaint: Content-Aware Image Retargeting for Any Ratio

Feihong Shen, Chao Li, Yifeng Geng, Yongjian Deng, Hao Chen

We combine semantic seam carving, adaptive repainting, and image enlargement to preserve salient content across challenging target ratios.

Paper Code

Information Fusion 2025JCR Q1

Modality-Balanced Contrastive Learning for RGB-D Salient Object Detection

Hao Chen, Feihong Shen, Lichuang Zhang

A contrastive learning framework that balances RGB and depth representation learning and transfers to RGB-D and RGB-T perception tasks.

Paper

05 / Side projects

Small tools, practical ideas.

Campus automation scripts Marvel Knowledge Graph