Dale Decatur

CS PhD student @ Uchicago

[Dale Decatur @DecaturDale]

Excited to share our #ICCV2025 work Reusing Computation in Text-to-Image Diffusion for Efficient Generation of Image Sets!

Our method generates sets of images using significantly less compute than standard diffusion.

📎https://t.co/MTYq2URf7T

1/

[Dale Decatur @DecaturDale]

Come check out our #CVPR2026 paper matching semantic segmentations between images and 3D shapes!

[Itai Lang @ItaiLang]

📢 Thrilled to present our work Best Segmentation Buddies 🤝 at #CVPR2026 this week!

Best Segmentation Buddies is a zero-shot method that matches semantic parts between images in the wild and untextured 3D shapes, even across completely different domains! 🦉✈️

1/6

[Richard Liu @allthatissolids]

Very excited to be presenting our highlight paper “Deep Feature Deformation Weights” at #CVPR2026!

TLDR: Image features make surprisingly good deformation weights. Our feature distillation is the best in the biz.

ExHall A&F 555
Poster Session 4
Website: https://t.co/Phh3WWV2QX

[Richard Liu @allthatissolids]

Going to be presenting our oral work WIR3D at @ICCVConference today! Please come check it out!
1:45 pm HST @ Exhibit Hall 3

We will also be #384 at poster session 4 afterwards if you want to chat.

[Dale Decatur @DecaturDale]

This work was completed in collaboration with @thibaultgroueix, @toomanyyifans, @RanaHanocka, Vladimir G. Kim, and Matheus Gadelha. Check out our #ICCV2025 poster #153 today during Poster Session #4 from 2:45-4:45 HST!

Paper: https://t.co/uy5ilrKMsl

7/

[Dale Decatur @DecaturDale]

Excited to share our #ICCV2025 work Reusing Computation in Text-to-Image Diffusion for Efficient Generation of Image Sets!

Our method generates sets of images using significantly less compute than standard diffusion.

📎https://t.co/MTYq2URf7T

1/

[Dale Decatur @DecaturDale]

While our method can be applied to general purpose image generation, our method achieves the most dramatic savings (saves >75% vs standard diffusion) when examples are structurally similar. Some applications of this are style variation, subject variation, and virtual try-on.

6/

[Rana Hanocka @RanaHanocka]

Our work “Geometry in Style” will be presented at #CVPR2025 on Sunday at 4pm in ExHall D, poster 219. Drop by and say hi!

Our technique is capable of performing expressive text-driven deformations that preserve the input shape identity.
1/

[Rana Hanocka @RanaHanocka]

3DL is recruiting PhD students to start in Fall 2025!

A thread👇highlighting some of our recent works, and what we are excited to explore next 🚀

Apply by Dec. 16; more info: https://t.co/WyvimLwjXe 1/

[Rana Hanocka @RanaHanocka]

📢 Our work iSeg will be presented at #SIGGRAPHAsia2024 today at 340PM Hall B5 (1) - come say hi!

iSeg is a new interactive segmentation technique for 3D shapes, producing fine-grained customized segmentations based on user clicks 👆

Project page: https://t.co/ZQiYoYr77s

1/

[Rana Hanocka @RanaHanocka]

Excited to share MeshUp ⬆️, a framework for deforming a mesh into a blend of various target concepts! MeshUp offers control over the influence of each concept by changing the associated weights. Project page: https://t.co/vYkXCjH63R

🧵1/

[Rana Hanocka @RanaHanocka]

Come check out TEDi today from 11:05-11:15 in mile high 4 https://t.co/iWwQvS3zpG #SIGGRAPH2024

[Cohere Labs @Cohere_Labs]

Tomorrow tune in with @DecaturDale as he presents "Text-Driven Localization and Local Editing of 3D Shapes" with our community-led Computer Vision group! See you there! 👨‍💻
Learn more: https://t.co/ELhQxyKmI8

[Dale Decatur @DecaturDale]

Come check out our #CVPR2024 poster for 3D Paintbrush 🎨🖌️ on 3D text-driven local editing! Happy to chat more at poster #420 in today’s 10:30 session.

[Rana Hanocka @RanaHanocka]

Excited to share 🎨🖌️ 3D Paintbrush - a method for generating local stylized textures on meshes using text as input! Our method predicts a localization map & a highly detailed texture map which conforms to it https://t.co/Fu9IhMpNRL (1/3)

[Dale Decatur @DecaturDale]

@_jasonliu_ While this work focused on local edits, you can use our CSD loss to perform global texturing as well (such as giving tiger stripes to the entire object)!

[Dale Decatur @DecaturDale]

Our 3D Paintbrush code is now live: https://t.co/8YUV9XgbeW. Check out our new Cascaded Score Distillation (CSD) loss! If you are already using SDS, you can swap it for CSD to enable higher resolution supervision from cascaded diffusion models. (1/3)

[Dale Decatur @DecaturDale]

The idea behind CSD is to leverage all stages of a cascaded diffusion model for supervision. SDS only uses the base (lowest res) stage of a cascaded diffusion model, while our CSD extends SDS to include the super resolution stages as well. (3/3)

[Dale Decatur @DecaturDale]

We use CSD to texture meshes, but we speculate that CSD will work on other representations such as NeRFs, images, and more. We created a simple example that applies CSD to image generation & editing: https://t.co/daR8Us7Fy4. (2/3)