AI with Papers - Artificial Intelligence & Deep Learning
15K subscribers
96 photos
238 videos
11 files
1.27K links
All the AI with papers. Every day fresh updates on Deep Learning, Machine Learning, and Computer Vision (with Papers).

Curated by Alessandro Ferrari | https://www.linkedin.com/in/visionarynet/
Download Telegram
This media is not supported in your browser
VIEW IN TELEGRAM
🦜 ColorDiffuser: Text-to-Video Colorization 🦜

👉HK University unveils ColorDiffuser: adapting pre-trained text-to-image latent diffusion model for video colorization

😎Review https://t.ly/XGv_
😎Paper arxiv.org/pdf/2306.01732.pdf
😎Project colordiffuser.github.io/
😎Code github.com/ColorDiffuser/ColorDiffuser
🤯82🤩1
This media is not supported in your browser
VIEW IN TELEGRAM
🌻 Extending Mona Lisa with AI 🌻

👉 A guy on Reddit extends Mona Lisa Painting with #Photoshop AI. The result is surprising.

😎More https://t.ly/j_2r
🤯20👍5🤩4🔥3😱2🤣21
This media is not supported in your browser
VIEW IN TELEGRAM
🏸 Segment Anything in HQ 🏸

👉HQ-SAM: SAM with the ability to accurately segment objects, maintaining promptable design, efficiency, zero-shot generalizability

😎Review https://t.ly/GxX5B
😎Paper arxiv.org/pdf/2306.01567.pdf
😎Models github.com/SysCV/SAM-HQ
🔥18👍4🤯1😱1😍1
This media is not supported in your browser
VIEW IN TELEGRAM
🌈 Track Everything Everywhere 🌈

👉#Google unveils OmniMotion: full-length motion tracking for every pixel in every frame of video.

😎Review https://t.ly/Krvw
😎Paper arxiv.org/pdf/2306.05422.pdf
😎Project omnimotion.github.io/
😎Demo omnimotion.github.io/#interactive_demo
😎Code github.com/qianqianwang68/omnimotion
🔥235🤯3🤩1💩1
This media is not supported in your browser
VIEW IN TELEGRAM
👁️ Scene Five: Through Her Eyes 👁️

👉 #3D scene reconstruction of what a person is observing using only the reflections of their eyes

😎Review https://t.ly/uBO6
😎Paper arxiv.org/pdf/2306.09348.pdf
😎Project https://world-from-eyes.github.io/
🤯28🔥12💩2🤩1
This media is not supported in your browser
VIEW IN TELEGRAM
🫣 Text-Guided Adversarial Makeup 🫣

👉Novel facial privacy protection via adversarial latent codes. Makeup vs Face Recognition.

😎Review https://t.ly/pBCP
😎Paper arxiv.org/pdf/2306.10008.pdf
😎Code github.com/fahadshamshad/Clip2Protect
6👍1🔥1🥰1💩1
Media is too big
VIEW IN TELEGRAM
🦷 Few-Shot Geometry-Aware Keypoints 🦷

👉UBC (+Flawless AI) unveils the new SOTA in semantic keypoints localization. Suitable for faces, animals, cars, mouth, teeth & more

😎Review https://t.ly/-0qN
😎Paper arxiv.org/pdf/2303.17216.pdf
😎Project xingzhehe.github.io/FewShot3DKP/
🤯10👍422👏2🤩2🔥1
This media is not supported in your browser
VIEW IN TELEGRAM
🚔 Fooling Neural Forensic Classifiers 🚔

👉Adversarial faces able to fool the forensic classifiers, while remaining undetectable by humans

😎Review https://t.ly/33Cc
😎Paper arxiv.org/pdf/2306.13091.pdf
😎Project koushiksrivats.github.io/face_attribute_attack
😎Code github.com/koushiksrivats/face_attribute_attack
😢64👏2😱2🍾2👍1🤯1😍1
panohead_overview-min.gif
24.3 MB
🍥 PanoHead: 3D Full-Head Synthesis 🍥

👉#ByteDance (+UW-M) unveils PanoHead: 360◦ view-consistent portraits from a single-view image

😎Review https://t.ly/MrLNR
😎Paper arxiv.org/pdf/2303.13071.pdf
😎Project sizhean.github.io/panohead
😎Code github.com/sizhean/panohead
🔥74🤯3😱1
This media is not supported in your browser
VIEW IN TELEGRAM
🔮SAM-PT: Segment Anything+Tracking🔮

👉SAM-PT is the first method to utilize sparse point propagation for Video Object Segmentation (VOS).

😎Review https://t.ly/QLMG
😎Paper arxiv.org/pdf/2307.01197.pdf
😎Project www.vis.xyz/pub/sam-pt/
😎Code github.com/SysCV/sam-pt
🔥147🤯3👍1😱1
This media is not supported in your browser
VIEW IN TELEGRAM
🪩 DISCO: Human Dance Generation 🪩

👉NTU (+ #Microsoft) unveils DISCO: a big step towards the Human Dance Generation.

😎Review https://t.ly/cNGX
😎Paper arxiv.org/pdf/2307.00040.pdf
😎Project disco-dance.github.io/
😎Code github.com/Wangt-CN/DisCo
🔥13🥰4😍21👍1🍾1
This media is not supported in your browser
VIEW IN TELEGRAM
🛣️ STAR.: 3D-tracking w/ attention paradigm 🛣️

👉#Mercedes STAR: e2e 3D object tracking that follows the tracking-by-attention paradigm

😎Review https://t.ly/JoGj
😎Paper arxiv.org/pdf/2306.17602.pdf
😎Project simondoll.github.io/publications/star_track
👍14🔥1🥰1👏1
This media is not supported in your browser
VIEW IN TELEGRAM
🍡 Text2Cinemagraphs: Cinemagraph from text 🍡

👉CMU (+ #Snap) unveils a fully automated method for creating cinemagraphs from text descriptions

😎Review https://t.ly/BwZs6
😎Paper arxiv.org/pdf/2307.03190.pdf
😎Project text2cinemagraph.github.io/website
😎Code github.com/text2cinemagraph/text2cinemagraph
12🤯3😱1🤩1
This media is not supported in your browser
VIEW IN TELEGRAM
🔥Test-Time Training on fire 🔥

👉Extending the TTT to the streaming setting. Suitable for Panoptic, Instance & Colorization.

😎Review https://t.ly/eZYA
😎Paper arxiv.org/pdf/2307.05014.pdf
😎Project https://video-ttt.github.io/
😎Code github.com/renwang435/video-ttt-release
🔥10👍31🤯1
This media is not supported in your browser
VIEW IN TELEGRAM
🃏 Deepfake via casual self-scan 🃏

👉TAU presents a novel approach to reenact an ID using only a casual self-scan

😎Review https://t.ly/9T8Wi
😎Paper arxiv.org/pdf/2307.06307.pdf
😎Project arielazary.github.io/PGR
🤯7👍65🔥1😱1
This media is not supported in your browser
VIEW IN TELEGRAM
🎪 Extreme Human Pose Estimation 🎪

👉RePoGen: novel synthetic data generator of extreme/realistic poses of humans

😎Review https://t.ly/ecBvM
😎Paper arxiv.org/pdf/2307.06737.pdf
😎Project mirapurkrabek.github.io/RePoGen-paper
😎Code github.com/MiraPurkrabek/RePoGen
🔥12👍2👏1🤯1😱1
This media is not supported in your browser
VIEW IN TELEGRAM
🧯Neural Focal Modulation VAR🧯

👉A novel architecture for video recognition that models both local/global context

😎Review https://t.ly/rF_fk
😎Paper arxiv.org/pdf/2307.06947.pdf
😎Project talalwasim.github.io/Video-FocalNets
😎Code github.com/TalalWasim/Video-FocalNets
🔥81👏1🤩1