🦧 #Nvidia Describe Anything 🦧

👉Nvidia unveils Describe Anything Model (DAM) the new SOTA in generating detailed descriptions for user-specified regions in images/videos, marked by points, boxes, scribbles, or masks. Repo under Apache, Dataset available and live demo on 🤗

👉Review https://t.ly/la4JD
👉Paper https://lnkd.in/dZh82xtV
👉Project https://lnkd.in/dcv9V2ZF
👉Repo https://lnkd.in/dJB9Ehtb
🤗Demo https://lnkd.in/dXDb2MWU

📍Moving Points -> Depth📍

👉KAIST & Adobe propose Seurat, a novel method that infers relative depth by examining the spatial relationships and temporal evolution of a set of tracked 2D trajectories (via off-the-shelf point tracking models). Repo & Demo to be released💙

👉Review https://t.ly/qA2P5
👉Paper https://lnkd.in/dpXDaQtM
👉Project https://lnkd.in/d9qWYsjP
👉Repo https://lnkd.in/dZEMDiJh

🌼SOTA Textured 3D-Guided VTON🌼

👉#ALIBABA unveils 3DV-TON, a novel diffusion model for HQ and temporally consistent video. Generating animatable textured 3D meshes as explicit frame-level guidance, alleviating the issue of models over-focusing on appearance fidelity at the expanse of motion coherence. Code & benchmark to be released💙

👉Review https://t.ly/0tjdC
👉Paper https://lnkd.in/dFseYSXz
👉Project https://lnkd.in/djtqzrzs
👉Repo TBA

🍏#Nvidia Dynamic Pose 🍏

👉Nvidia unveils DynPose-100K, the largest dataset of dynamic Internet videos annotated with camera poses. Dataset released under Nvidia license💙

👉Review https://t.ly/wrcb0
👉Paper https://lnkd.in/dycGjAyy
👉Project https://lnkd.in/dDZ2Ej_Q
🤗Data https://lnkd.in/d8yUSB7m

🔥 S3MOT: SOTA 3D MOT 🔥

👉S3MOT: Selective-State-Space model-based MOT that efficiently infers 3D motion and object associations from 2D images through three core components. New SOTA on KITTI with 76.86 HOTA at 31 FPS! Code & Weights to be released under MIT license💙

👉Review https://t.ly/H_JPv
👉Paper https://arxiv.org/pdf/2504.18068
👉Repo https://github.com/bytepioneerX/s3mot

🔥 Diffusion Model <-> Depth 🔥

👉ETH & CMU on how to turn a single-image latent diffusion model (LDM) into the SOTA video depth estimator: video depth without video models. Repo released under Apache 2.0 and HF demo available💙

👉Review https://t.ly/sP9ma
👉Paper arxiv.org/pdf/2411.19189
👉Project rollingdepth.github.io/
👉Repo github.com/prs-eth/rollingdepth
🤗Demo huggingface.co/spaces/prs-eth/rollingdepthhttps://t.ly/sP9ma

🩷Dance vs. #ComputerVision🩷

👉The Saint-Etienne university proposed a new 3D human body pose estimation pipeline to deal with dance analysis. Project page w/ results and interactive demo released💙

👉Review https://t.ly/JEdM3
👉Paper arxiv.org/pdf/2505.07249
👉Project https://lnkd.in/dD5dsMv5

🧞‍♀️GENMO: Generalist Human Motion 🧞‍♀️

👉#Nvidia presents GENMO, a unified Generalist Model for Human Motion that bridges motion estimation and generation in a single framework. Conditioning on videos, 2D keypoints, text, music, and 3D keyframes. No code at the moment🥲

👉Review https://t.ly/Q5T_Y
👉Paper https://lnkd.in/ds36BY49
👉Project https://lnkd.in/dAYHhuFU

Dear friends,
I’m truly sorry for being away from the group for so long. I know: no updates so far while AI is running faster than speed of light.

I’m going through a very difficult time in my life and I need some space to heal. This spare-time project (but important for a lot of people here) needs energy and commitment I don’t have right now. I’m sorry, be patient. I’ll be back.

Love u all,
Alessandro.

Hi everybody,
I took a few weeks to take a breath from a lot of stuff, I dedicated all my mental energy to keep working and I dedicated all my spare time to take care of myself. Despite I'm still not ok (BTW, my health was/is always good), I feel it's time to come back and support this wonderful community in this journey. I feel the responsibility of that, time to get in the ring.

I'm very sorry for being out so long, but sometime life hits really hard. I got an incredible support from unknown people from all around the world. It's amazing.

Thanks again, you rock!
Alessandro.

🦖 DINOv3 is out 🦖

👉#Meta unveils DINOv3! A novel foundation model outperforming the previous SOTAs in computer vision. Code & weights released under DINOv3 License💙

👉Review https://t.ly/-S3ZL
👉Paper https://t.ly/ervOT
👉Project https://lnkd.in/dHFf3esd
👉Repo https://lnkd.in/dPxhDxAq
🤗HF https://lnkd.in/dWGudY2i

🤖 Impact of SuperHuman AI 🤖

👉The NoProfit AI Futures Project unveils a (dystopic) scenario about what super-AI might look like. Forecast from today to the bio-engineered human-like creatures. A fascinating speculation of the future with the "slow-down" and "race" scenarios. Enjoy 💙

👉Review https://t.ly/EgmfJ
👉Project https://ai-2027.com/

🏓TOTNet: Occlusion-aware Tracking🏓

👉TOTNet: novel Temporal Occlusion Tracking Network that leverages 3D-convs, visibility-weighted loss, & occlusion augmentation to improve performance under occlusions. Code & Data under MIT💙

👉Review https://t.ly/Q0jAf
👉Paper https://lnkd.in/dUYsa-GC
👉Repo https://lnkd.in/d3QGUHYb

🔀Feed-Forward 4D video🔀

👉4DNeX is the first feed-forward framework for generating 4D scene representations from a single image by fine-tuning diffusion model. HQ dynamic pt-clouds & downstream tasks such as novel-view video synthesis with strong generalizability. Code/Data announced 💙

👉Review https://t.ly/SpkD-
👉Paper arxiv.org/pdf/2508.13154
👉Project https://4dnex.github.io/
👉Repo github.com/3DTopia/4DNeX
👉Data https://lnkd.in/dh4_3Ghf
👉Demo https://lnkd.in/dztyzwgg

🌈DAViD: Synthetic Depth-Normal-Segmentation🌈

👉#Microsoft's DAViD: 100% synthetic dataset/models for human Depth, Normals & Segmentation. Dataset available, models & runtime under MIT💙

👉Review https://t.ly/-SlO_
👉Paper https://lnkd.in/eCmMXpTg
👉Project https://lnkd.in/eurCSWkm
👉Repo https://lnkd.in/e7PWFgP2

👠 OmniTry: Virtual Try-On Anything 👠

👉OmniTry: unified framework that extends VTON beyond garment to encompass any wearable objects (jewelries, accessories, etc.) in mask-free setting. Weights, HF demo & benchmark released💙

👉Review https://t.ly/wMBGQ
👉Paper https://lnkd.in/dQe9MchS
👉Project https://omnitry.github.io/
👉Repo https://lnkd.in/d3QwAXY2
🤗Demo https://lnkd.in/duUcZpVA

📡 ROVR Open Dataset is out 📡

👉A novel large-scale open 3D dataset for autonomous driving, robotics, and 4D perception tasks. To be released for academic (for free) & commercial💙

👉Review https://t.ly/iDcvg
👉Paper https://arxiv.org/pdf/2508.13977
👉Project https://xiandaguo.net/ROVR-Open-Dataset

🧉 YOPO: SOTA 9-DoF Pose🧉

👉Pit In Co. unveils YOPO, a novel single-stage, query-based framework that treats category-level 9-DoF estimation as a natural extension of 2D detection. A practical solution for mono-RGB, category-level, multi-obj pose estimation. Code & models announced (coming)💙

👉Review https://t.ly/cf_Cl
👉Paper https://arxiv.org/pdf/2508.14965
👉Project mikigom.github.io/YOPO-project-page/
👉Repo TBA

🔬Intern-S1: SOTA MM-MoE 🔬

👉InternS1: a MM-MoE with 28B activated / 241b total parameters, continually pre-trained on 5T tokens, including 2.5T+ tokens from scientific domains. New SOTA for professional tasks, such as molecular synthesis planning, reaction condition prediction, etc. Models available under Apache 2.0💙

👉Review https://t.ly/3l5UW
👉Paper arxiv.org/pdf/2508.15763
👉Repo github.com/InternLM/Intern-S1
🤗HF huggingface.co/internlm/Intern-S1