Zezhou Cheng

Assistant Professor @CS_UVA; PostDoc @Caltech; Ph.D. @UMass CICS; Interested in computer vision and its applications.

[Zezhou Cheng @ZezhouCheng]

Thanks @_akhaliq for sharing our work!

[AK @_akhaliq]

OmniShotCut

Holistic Relational Shot Boundary Detection with Shot-Query Transformer

paper: https://t.co/dNBF7bUeE8

[Boyang Wang @BoyangWang7]

Introducing OmniShotCut, a sensitive and more informative SoTA on the Shot Boundary Detection.
OmniShotCut can detect shot changes of the video in diverse sources (anime, vlog, game, shorts, sports, screen recording, etc.), and recognize Sudden Jump and Transitions.

[Wildminder @wildmindai]

OmniShotCut.
Relational shot boundary detection via Shot-Query Transformer.
- detects abrupt cuts and gradual transitions;
- automatic video editing and indexing;
SOTA performance

https://t.co/rMsFr5D2lT

[Zezhou Cheng @ZezhouCheng]

Exciting to see progress in this line of work, and glad our LabelAny3D (NeurIPS 2025) is being adopted!

[DailyPapers @HuggingPapers]

AllenAI releases WildDet3D for promptable 3D detection in the wild

Understands objects in 3D from a single photo—predicting position, size, and orientation. Supports text, point, and box prompts, with optional depth integration.

[Zezhou Cheng @ZezhouCheng]

@haoshu_fang @umdcs Congratulations, Hao-Shu! What an achievement! Welcome to visit us at UVA anytime.

[Zhenjun Zhao @zhenjun_zhao]

SAB3R: Semantic-Augmented Backbone in 3D Reconstruction

@ChenXuweiyi, @TianX_ia, @6SihanXu, @jed_yang, Joyce Chai, @ZezhouCheng

tl;dr: MASt3R+distillation->open-vocabulary segmentation+3D reconstruction

https://t.co/rCVX8kZfSU

[Boyang Wang @BoyangWang7]

Thanks @_akhaliq for sharing!
The homepage can be found at: https://t.co/lAZ3IfWpY0

Thank you to all my coauthors: @ChenXuweiyi Matheus Gadelha @ZezhouCheng
Research Work at UVA!
Special thanks @ChenXuweiyi for bridging this opportunity.

[Boyang Wang @BoyangWang7]

Introducing Frame In-N-Out: Unbounded Controllable Image-to-Video Generation

Can we enable video generation to capture a wider and more imaginative world that is not confined by the spatial boundaries of the initial frame?

[Jin Yao @Jin_Yao_]

🚀 Thrilled to share our latest work: Open Vocabulary Monocular 3D Object Detection

💡 Our method OVMono3D-LIFT can detect and localize objects in 3D from a single RGB image for any object category!

🌟 Results on In-the-Wild COCO images:

A 🧵:

#ComputerVision #3DDetection

[Zezhou Cheng @ZezhouCheng]

Very excited to share this work from my awesome PhD student, @ChenXuweiyi! We’d love to hear any feedback, thoughts or questions!

[Xuweiyi Chen @ChenXuweiyi]

🚀 Excited to share our latest paper: “Learning 3D Representations from Procedural 3D Programs”

We explore self-supervised learning of 3D representations using procedurally generated shapes, with no reliance on human-designed 3D datasets. We found that Self-supervised 3D representation learning from procedural 3D programs performs on par with learning from ShapeNet across various downstream 3D tasks. Both outperforms training from scratch by a large margin.
👇
📄 https://t.co/0ri6vjgPsU
💾 Dataset: https://t.co/Ynkiaivwi5
🖥️ Code: https://t.co/s4O2b7BryC
#3DRepresentation #PointClouds #SelfSupervisedLearning

[Oindrila Saha @oindy_saha]

We have open-sourced the code and data for "Improved Zero-Shot Classification by Adapting VLMs with Text Descriptions" @CVPR at: https://t.co/imdgtl5nwt
Also, find the poster and video links at the GitHub repo!

[Zezhou Cheng @ZezhouCheng]

R.I.P

[Yi Ma @YiMaTweets]

Today, I am deeply shocked and saddened by the news that my long-time dear friend, professor Xiaoou Tang of CUHK, advisor to Dr. Kaiming He and my predecessor at MSRA, peacefully passed away at home yesterday. I just had dinner with him last Wednesday in Shanghai! RIP... 🙏🙏🙏

[Manning College of Information & Computer Sciences @manningcics]

With colleagues from @umassengin, @MajiSubhransu + PhD students Gustavo Perez, Aaron Sun and @ZezhouCheng have developed ZeoNet, a #DeepLearning framework using #ComputerVision techniques to predict the separation performance of zeolite materials. Article: https://t.co/fFSGIFoOwp

[Subhransu Maji @MajiSubhransu]

@ClimateChangeAI has an excellent post on our work on modeling zeolites. Their 3D structures are predictive of their chemical separation performance, and can be modeled very well using 3D ConvNets.

[Rogério Guimarães @rogerioagjr]

We're excited to share our latest work! We achieve SOTA results in segmentation, detection, and depth estimation, in single and cross-domain, by exploiting image-aligned text prompts in a pretrained diffusion backbone repurposed for vision tasks.

See https://t.co/fGI2UfvJwS
🧵👇

[Zezhou Cheng @ZezhouCheng]

@georgiagkioxari @MajiSubhransu @jampani_varun @HuaizuJiang @gadelha_m Thanks, Georgia!! Looking forward to seeing you there soon!

[Zezhou Cheng @ZezhouCheng]

I am eagerly looking forward to my next journey at Caltech w/ @georgiagkioxari and UVA! Massive thanks to my awesome advisor @MajiSubhransu and committee members Dan, Erik, and @jampani_varun! Huge thanks to @HuaizuJiang, @gadelha_m, and many other friends for their support!

[Subhransu Maji @MajiSubhransu]

Huge congrats to @ZezhouCheng for successfully defending his PhD thesis! He will be starting as an asst. professor at @UVA after postdoc at @Caltech w/ @georgiagkioxari. Prospective students interested in few-shot learning and 3D shape understanding def. apply to work with him.

[Zezhou Cheng @ZezhouCheng]

@gadelha_m @georgiagkioxari @MajiSubhransu @jampani_varun @HuaizuJiang Hahaha, indeed! Looking forward to more collaborations in the future!

[Zezhou Cheng @ZezhouCheng]

@HuaizuJiang @georgiagkioxari @MajiSubhransu @jampani_varun @gadelha_m Thank you so much, Huaizu!

[Zezhou Cheng @ZezhouCheng]

@kamalgupta09 @georgiagkioxari @MajiSubhransu @jampani_varun @HuaizuJiang @gadelha_m Thanks, Kamal!!