Subin Kim

Incoming Applied Scientist @ Adobe

[Changyeon Kim @cykim1006]

Heading to #ICLR2026 in Rio 🇧🇷 this week to present DEAS!

📍 Poster session: Fri, April 24, 10:30 AM – 1:00 PM
📌 Location: Pavilion 3 P3-#1308

Happy to chat about robot learning, VLAs, and RL!
I’ll be on the job market soon and looking for exciting opportunities!
#ICLR

[June Suk Choi @june_suk_choi]

Excited to share Adaptive Low-Pass Guidance (ALG): a simple training-free, drop-in fix that brings dynamic motion back to Image-to-Video models! Demo videos, paper, & code below!
https://t.co/4NzYDfCFSb (🧵 1/7)

[Younggyo Seo @younggyoseo]

Excited to present FastTD3: a simple, fast, and capable off-policy RL algorithm for humanoid control -- with an open-source code to run your own humanoid RL experiments in no time!

Thread below 🧵

[Xiaolong Wang @xiaolonw]

Cool work from @JitengMu on image editing! This looks like the future of Photoshop!

Just select a few patches and move it, and boom, you get the edited picture! This enables the new control ability of the diffusion model: Only change the part you want to change.
https://t.co/x2xeUy6GaW

[Radamés Ajna @radamar]

Official Demos:
Hyper-SDXL-1Step-T2I https://t.co/BWPyYACp4t

Hyper-SD15-Scribble
https://t.co/NSWR5P4Uee

Unoffical Demos:
InstantStyle + Hyper SD1.5 (not great but superfast)
https://t.co/VDQQ8FAj8X
InstantStyle + Hyper SDXL
https://t.co/RjWmkiGCFP

[Jaeho Lee @jaeho_lee_]

Suppose that we train two INRs: One for a natural image, and another for its pixel-shuffled version.

Which INR would fit faster?

Expected: Natural Image
Reality: Pixel-permuted image 🤯
(under some conditions)

We look closer into when & why this happens in our #CVPR2024 oral.

[Sihyun Yu @sihyun_yu]

We are organizing INRs for vision workshop at CVPR 2024! Please submit your awesome work related to INRs/neural fields 🙂

[Abhinav Shrivastava @abhi2610]

Call for Papers: #INRV2024 Workshop on Implicit Neural Representation for Vision @ #CVPR2024! Topics: Compression, Representation using INR’s for images, audio, video & more! Ddl: 3/31. Submit now! @CVPR
Website: https://t.co/380CMq5Ouc
Submission Link: https://t.co/KwkGqyKjhy

[Subin Kim @subin_kim____]

@srbhsxn @JunhwaHur @DeqingSun @fleet_dj Thanks for greate work. When the code will be released?

[Jihoon Tack @jihoontack]

🙌 At #NeurIPS to present two papers
Wed. 10AM: Efficient meta-learning / 5PM: Improving MAE with meta-learning.

I am working on continual learning/efficiency/AutoML for LLMs! DM me if you are interested in chatting
👀 Also looking for an internship, welcome any recommendations

[Changyeon Kim @cykim1006]

Excited to share Adaptive Return-conditioned Policy (ARP): a return-conditioned policy utilizing adaptive multimodal reward from pre-trained CLIP encoders!

ARP can mitigate goal misgeneralization and execute unseen text instructions!

https://t.co/4912g4XalS

🧵👇
1/N

[nama @aman_gif]

colab for clip guided gaussian splatting: https://t.co/KcQLNsofzK cc @GKopanas :)

[International Conference on 3D Vision (3DV) @3DVconf]

Best Paper Award Honorable Mention at #3DV2019

Learned Multi-View Texture Super-resolution

Audrey Richard; Ian Cherabier; Martin R. Oswald; Vagia Tsiminaki; Marc Pollefeys @mapo1; Konrad Schindler

#TBThursday #3DV2024

[Junsu Kim @JunsuKim97]

I will present Multi-View Masked World Models (MV-MWM) for visual robotic manipulation.

Please visit #ICML2023 poster session at 7/25 (Tue) 2:00-3:30pm!

[Tanmay Gupta @tanmay2099]

Imagine a 2D image serving as a window to a 3D world that you could reach into, manipulate objects, and see changes reflected in the image.

In our new OBJect 3DIT work, we edit images in this 3D-aware fashion while only operating in the pixel space!

🧵

[Zhenjun Zhao @zhenjun_zhao]

ResMatch: Residual Attention Learning for Local Feature Matching

Yuxin Deng, Jiayi Ma

tl;dr: descriptor similarity->cross-attention; relative position->self-attention

https://t.co/wwbLABbI3S

[AK @_akhaliq]

Collaborative Score Distillation for Consistent Visual Synthesis

paper page: https://t.co/uMZnQUmgO2

Generative priors of large-scale text-to-image diffusion models enable a wide range of new generation and editing applications on diverse visual modalities. However, when adapting these priors to complex visual modalities, often represented as multiple images (e.g., video), achieving consistency across a set of images is challenging. In this paper, we address this challenge with a novel method, Collaborative Score Distillation (CSD). CSD is based on the Stein Variational Gradient Descent (SVGD). Specifically, we propose to consider multiple samples as "particles" in the SVGD update and combine their score functions to distill generative priors over a set of images synchronously. Thus, CSD facilitates seamless integration of information across 2D images, leading to a consistent visual synthesis across multiple samples. We show the effectiveness of CSD in a variety of tasks, encompassing the visual editing of panorama images, videos, and 3D scenes. Our results underline the competency of CSD as a versatile method for enhancing inter-sample consistency, thereby broadening the applicability of text-to-image diffusion models.

[Jihoon Tack @jihoontack]

🤔 Learning large-scale neural fields (NFs) requires a huge amount of memory and time to train…

💡I’m excited to share “Learning Large-scale Neural Fields via Context Pruned Meta-Learning”.

Paper: https://t.co/usIAAwtVFh

[Davis Blalock @davisblalock]

"JPEG Compressed Images Can Bypass Protections Against AI Editing"

Since neural nets are vulnerable to adversarial perturbations, you might hope that you could modify your image to resist AI image editing. But... [1/2]

[AK @_akhaliq]

ELITE: Encoding Visual Concepts into Textual Embeddings for Customized Text-to-Image Generation code release is out

github: https://t.co/mhijcWUVc6
@Gradio demo: https://t.co/fHKEcMwNFU