🔥 Is ChatGPT (close to) an AGI? 🔥

👉According to Microsoft #ChatGPT is an early (yet still incomplete) version of an artificial general intelligence system. Math, coding, vision, medicine, law, psychology and more.

👉Flame here: https://bit.ly/3K3Q1XC 🤣

🪲Neural 4K Color Style Transfer🪲

👉#Adobe 4K color style transfer: no artifacts, low memory & super-speed!

😎Review https://bit.ly/3TNKtUE
😎Paper arxiv.org/pdf/2303.13511.pdf
😎Project zhkkke.github.io/NeuralPreset
😎Code github.com/ZHKKKe/NeuralPreset
😎Demo zhkkke.github.io/NeuralPreset/#/editor

🥦 ReBotNet: Neural Enhancement 🥦

👉#Google unveils ReBotNet, novel real-time video enhancement for live video calls & streams

😎Review https://bit.ly/3z8oqhG
😎Paper arxiv.org/pdf/2303.13504.pdf
😎Project jeya-maria-jose.github.io/rebotnet-web

📊 Neural vs. Old School ML 📊

✅ The choice of Deep Learning-based AI vs. traditional ML may depend on the specific application and problem being addressed.

👉 What's your favorite approach? Please vote: https://bit.ly/3LQU774

🛎 Neural Tracking + 3D Reconstruction 🛎

👉The novel SOTA for real-time 6-DoF tracking of unknown objects, simultaneously performing neural 3D

😎Review https://bit.ly/3lHVoT2
😎Paper arxiv.org/pdf/2303.14158.pdf
😎Project bundlesdf.github.io/

🦒SOTA ALERT: Synthetic 3D Cities 🦒

👉UrbanGIRAFFE: novel SOTA generator of photorealistic #3D-aware urban image with full controllability

😎Review https://bit.ly/42ZdX5Q
😎Paper arxiv.org/pdf/2303.14167.pdf
😎Project lv3d.github.io/urbanGIRAFFE

🔥 "Segmenting Anything". CRAZY! 🔥

👉#Meta unveils a novel model and (1B+) dataset for neural segmentation 🤯

😎Review https://bit.ly/3nM2uXx
😎Paper https://bit.ly/43788DC
😎Project https://segment-anything.com
😎Code github.com/facebookresearch/segment-anything

💦 SegGPT: Segmenting Everything (In Context) 💦

👉BAAI unveils SegGPT, a generalist model for segmenting everything in context

😎Review https://bit.ly/3zFkUf2
😎Paper arxiv.org/pdf/2304.03284.pdf
😎Code github.com/baaivision/Painter
😎Demo huggingface.co/spaces/BAAI/SegGPT

👗DreamPose: Fashion I-2-V Diffusion👗

👉 Turning fashion photos into realistic videos via driving pose sequence

😎Review https://bit.ly/3AdNtAN
😎Paper arxiv.org/pdf/2304.06025.pdf
😎Code github.com/johannakarras/DreamPose
😎Project grail.cs.washington.edu/projects/dreampose

🥦 Zip-NeRF: the Anti-Aliasing NeRF 🥦

👉#Google unveils a novel version of NeRF able to fix the aliasing problem being 22x faster in training than SOTA.

😎Review https://bit.ly/3L1hZ6M
😎Paper arxiv.org/pdf/2304.06706.pdf
😎Project https://jonbarron.info/zipnerf

🔥 ALERT: Stable Diffusion XL is out! 🔥

👉SDXL the new generative AI by Stability.AI for images from text. Up to 1024x1024 resolution, for free.

😎More https://bit.ly/41wrh0j

🪬 META's Animated Drawings is out! 🪬

👉#META unveils an easy-to-use method for animating human-like figures drawn by children.

😎Review https://bit.ly/3mGeQQv
😎Paper arxiv.org/pdf/2303.12741.pdf
😎Project fairanimateddrawings.com

🌻DDS: diffusive text-based image editing🌻

👉Google unveils a novel text-based image editing for modifications of an input image towards a text description.

😎Review https://bit.ly/3L52UBl
😎Paper arxiv.org/pdf/2304.07090.pdf
😎Project delta-denoising-score.github.io

🪅 Inpaint Anything: Segmentation + Inpainting 🪅

👉Remove / Fill /Replace anything (also via prompt). "Inpaint Anything", a new paradigm of “clicking & filling"

😎Review https://bit.ly/43JNREE
😎Paper arxiv.org/pdf/2304.06790.pdf
😎Code github.com/geekyutao/Inpaint-Anything

Hi friends,
right now I'm flying to NY for a business trip!

👉 Is there anyone studying/working @NYU? I'd love to visit the campus and (eventually) attend to a few lessons about AI/CV/MATH on Monday (or this Friday)

Send me a DM -> @argovision

🔥 Track Anything: SAM-powered tracking 🔥

👉 SUSTech VIP Lab proposes TAM, a "novel" video tracker powered by SAM

😎Review https://bit.ly/44jwI4W
😎Paper arxiv.org/pdf/2304.11968.pdf
😎Code github.com/gaomingqi/Track-Anything

🌱 Segment Everything Everywhere 🌱

👉 Segmenting everything using visual/language prompts (BBs, scribbles, text & audio)

😎Review https://bit.ly/3LEiOmx
😎Paper arxiv.org/pdf/2304.06718.pdf
😎Demo huggingface.co/spaces/xdecoder/SEEM
😎Code github.com/UX-Decoder/Segment-Everything-Everywhere-All-At-Once

🦒 Look mom, I'm a giraffe 🦒

👉 A patent to transpose adversarial patches onto a knitted fabric. Be undetectable or associated with incorrect category such as "animal" (giraffe, zebra, etc)

😎 More: https://bit.ly/3LzjSGV

🐊 RelPose++: SOTA 6D from 2-8 pics 🐊

👉CMU unveils a novel neural method for 6D camera poses from only 2-8 images

😎Review https://bit.ly/42ioJ6K
😎Paper arxiv.org/pdf/2305.04926.pdf
😎Project amyxlase.github.io/relpose-plus-plus
😎Code github.com/amyxlase/relpose-plus-plus

🦕 6D Non-Prehensile Manipulation 🦕

👉#META (+CMU) unveils HACMan, novel 6D non-prehensile manipulation of objects

😎Review https://bit.ly/3NP1jl1
😎Paper arxiv.org/pdf/2305.03942.pdf
😎Project hacman-2023.github.io

🛸 Virtual Occlusions in #AR 🛸

👉Niantic (#pokemongo) on a novel approach for virtual assets to appear ‘sitting among’ the real world objects

😎Review https://bit.ly/3o04wn6
😎Paper arxiv.org/pdf/2305.07014.pdf
😎Project nianticlabs.github.io/implicit-depth
😎Code github.com/nianticlabs/implicit-depth