AI with Papers - Artificial Intelligence & Deep Learning

🔥 Depth Anything: the new SOTA 🔥 👉Depth Anything: the new SOTA in monocular depth estimation (MDE), trained with 1.5M labeled images and 62M+ unlabeled images jointly. It's the new SOTA! Authors: University of Hong Kong, #TikTok, Zhejiang Lab & Zhejiang University. Paper, project, code, models & HF-demo released! 👉Review https://t.ly/tCBwO 👉Paper https://lnkd.in/djx-9k2J 👉Project https://lnkd.in/dYetqZFa 👉Repo https://lnkd.in/d87CrUGv 👉Demo🤗 https://lnkd.in/dJhvKBep

😻 GARField: Group Anything 😻 👉 GARField is a novel approach for decomposing #3D scenes into a hierarchy of semantically meaningful groups from posed image inputs. 👉Review https://t.ly/6Hkeq 👉Paper https://lnkd.in/d28mfRcZ 👉Project https://lnkd.in/dzYdRNKy 👉Repo (coming) https://lnkd.in/d2VeRJCS

👽 One Model <-> All Segmentations 👽 👉 10+ different segmentation tasks in one framework, including image-level, video-level, interactive segmentation, & open-vocabulary segmentation. 👉Review https://t.ly/fywVz 👉Paper https://lnkd.in/dw3S4B74 👉Project https://lnkd.in/dzHT9v45 👉Repo https://lnkd.in/d6fDCnSp

🦠 XINC: Pixels to Neurons 🦠 👉eXplaining the Implicit Neural Canvas (XINC) from the University of Maryland, is a unified framework for explaining properties of INRs by examining the strength of each neuron’s contribution to each output pixel 👉Review https://t.ly/wwAmz 👉Paper arxiv.org/pdf/2401.10217.pdf 👉Project namithap10.github.io/xinc 👉Repo github.com/namithap10/xinc

🫒 AlphaGeometry: Olympiad-level AI 🫒 👉 Theorem prover for Euclidean plane geometry that sidesteps the need for human demonstrations by synthesizing millions of theorems and proofs across different levels of complexity 🤯 👉Review https://t.ly/2-Z7C 👉Paper https://lnkd.in/g3QkqwCE 👉Blog https://lnkd.in/ge-mpM7q 👉Repo https://lnkd.in/gHjwks_9

🔥🔥 Code is out 🔥🔥 Check the comments for the links ;)

💃Timeline Text-Driven Humans💃 👉Novel challenge: timeline control for text-driven motion synthesis of 3D Humans. 👉Review https://t.ly/HLm-N 👉Paper https://lnkd.in/esaR_M_9 👉Project https://lnkd.in/epCZDvFW 👉Repo coming

💥 Announcing #Py4Ai Conference💥 👉 Super proud to unveil #Py4AI, the newest conference dedicated to exploring the depths of Python & AI. Py4AI is a 1-day free event for Python and Artificial Intelligence developers. 𝐄𝐯𝐞𝐧𝐭 𝐃𝐞𝐭𝐚𝐢𝐥𝐬: ✅16th March 2024 (Saturday) - all day long! ✅Location: Pavia, at SEA Vision HQ ✅Seats: first 100 available now! ✅Live streaming: YES 🌍 ✅Cost: it's 100% FREE! We will investigate the latest trends, advancements, and challenges in Artificial Intelligence: NLP, MLOps, Computer Vision, Machine Learning & Generative AI! A scientific conference with international speakers. Only Science, Coding & professionals in AI. And beers, of course. 𝐓𝐡𝐞 𝐟𝐢𝐫𝐬𝐭 𝐛𝐚𝐭𝐜𝐡 𝐨𝐟 𝐬𝐩𝐞𝐚𝐤𝐞𝐫𝐬: 🚀Merve Noyan | ML Advocate Eng. Hugging Face 🤗 🚀Gabriele Lombardi | CTO ARGO Vision 🚀Amanda Cercas Curry | PostDoc. Università Bocconi 🚀Piero Savastano | Founder Cheshire Cat AI 🚀Francesco Zuppichini | Sr. ML engineer Zurich Insurance 🚀Andrea Palladino, PhD | Sr. Data Scientist 👉 More: https://www.linkedin.com/posts/visionarynet_py4ai-py4ai-python-activity-7152928716988243968-pOUn?utm_source=share&utm_medium=member_desktop

💊💊 AGNOSTIC Object Counting 💊💊 👉PseCo: combining SAM to segment all possible objects as mask proposals & CLIP to classify proposals to obtain accurate object counts. The new SOTA in both few-shot/zero-shot object counting/detection. 👉Discussion https://lnkd.in/dMgakzWm 👉Paper https://lnkd.in/dbzMXKWG 👉Repo https://lnkd.in/db9Q9Pse

🥛 Transparent Object Tracking 🥛 👉Trans2k: transparent object tracking dataset of 2,000+ sequences with 100,000+ images, annotated by bounding boxes & segmentation mask. 👉Review https://t.ly/mEI6O 👉Paper https://lnkd.in/dsudY3DB 👉Project https://lnkd.in/d48SSJJ3 👉TOB https://lnkd.in/dykBUNfC

🃏ReplaceAnything: demo is out!🃏 👉ReplaceAnything: ultra-high quality content replacement. The ultimate #AI solution for human, clothing & background replacement to change the e-commerce experience for vendors. 👉Review https://t.ly/FMyvf 👉Project https://lnkd.in/dcyZvP2b 👉ModelScope https://lnkd.in/dU4x4nE6 👉Hugging Face https://lnkd.in/dn3uXWgd 👉Empty report https://lnkd.in/dcuGXd6c 👉Paper coming?

🔥 #6D Foundation Pose 🔥 👉#Nvidia unveils FoundationPose, a novel (and unified) foundation model for 6D object pose estimation and tracking. 👉Review https://t.ly/HGd4h 👉Project https://lnkd.in/dPcnBKWm 👉Paper https://lnkd.in/dixn_iHZ 👉Code coming 🩷

🐤 MagicVideo-V2 Video Generation 🐤 👉#Bytedance announces a novel multi-stage pipeline capable of generating high-aesthetic videos from textual description 👉Review https://t.ly/zIq4v 👉Project https://lnkd.in/dKUrJPJd 👉Paper https://lnkd.in/dixnN-kU

🕍 En3D: Generative 3D Humans 🕍 👉#Alibaba unveils En3D: generative scheme for sculpting HQ 3D human avatars. Zero-shot 3D generative scheme capable of producing visually realistic, geometrically accurate and content-wise diverse 3D humans without relying on pre-existing 3D or 2D asset. 👉Review https://t.ly/nGmDK 👉Project menyifang.github.io/projects/En3D/index.html 👉Paper https://arxiv.org/pdf/2401.01173.pdf 👉Repo (coming) https://github.com/menyifang/En3D

👻 W2A: Avatar Behind Occlusions 👻 👉Neural rendering for occluded in-the-wild mono-videos. Decoupling scenes in occlusion, human, and background. 👉Review https://t.ly/8q__B 👉Paper https://arxiv.org/pdf/2401.00431.pdf 👉Project https://cs.stanford.edu/~xtiange/projects/wild2avatar

🈚 Seeing Through Occlusions 🈚 👉Princeton University unveils a novel NSF approach able to fuse a burst image capture into one high-resolution reconstruction and decompose it into transmission and obstruction layers. Tasks: seeing through occlusions, reflection suppression, shadow removal 👉Review https://t.ly/5jcIG 👉Project https://light.princeton.edu/publication/nsf/ 👉Paper https://arxiv.org/pdf/2312.14235.pdf 👉Repo https://github.com/princeton-computational-imaging/NSF

🪲UniRef++: Segment Every Reference🪲 👉 UniRef++ is a unified model for RIS, FSS, RVOS & VOS. Code available! 👉Review https://t.ly/OxtOx 👉Paper https://lnkd.in/eTrmDTK3 👉Repo https://lnkd.in/etfTm4Wq

🪮HAAR: Text-Driven Generative Hairstyles🪮 👉 HAAR: new strand-based generative model for #3D human hairstyles driven by textual input. 👉Review https://t.ly/L38iD 👉Project https://haar.is.tue.mpg.de/ 👉Paper https://arxiv.org/pdf/2312.11666.pdf 👉Repo coming

🎄UniSDF: Unifying Neural Representations🎄 👉UniSDF: novel general purpose 3D reconstruction for large complex scenes with reflections. SOTA on DTU, Shiny Blender, Mip-NeRF 360 and Ref-NeRF dataset. 👉Review https://t.ly/2QEul 👉Paper https://arxiv.org/pdf/2312.13285.pdf 👉Repo: No code :(

🚿 Event-Camera (1000 fps) Hands 🚿 👉Ev2Hands, the first method for the 3D reconstruction of two interacting hands from a single event camera. Code available. 👉Review https://t.ly/YpQpX 👉Paper https://arxiv.org/pdf/2312.14157.pdf 👉Project https://4dqv.mpi-inf.mpg.de/Ev2Hands 👉Repo https://github.com/Chris10M/Ev2Hands