🔥 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

🍿 De-Aging Harrison Ford via SD 🍿

👉Stable Diffusion for Hollywood: preview of the next autotune of entertainment industry. A discussion👇

😎 More: https://bit.ly/41EzaQK

🪰 #3D Auto-Reconstruction 🪰

👉AutoRecon: automated discovery & reconstruction of objects from multi-view pics.

😎Review https://bit.ly/3MxI0f4
😎Paper arxiv.org/pdf/2305.08810.pdf
😎Project zju3dv.github.io/autorecon/
😎Code github.com/zju3dv/AutoRecon

👚 Multi-Layered 3D Garments Animation 👚

👉S-Lab unveils LayersNet: animating multi-layered garments driven by various external forces, such as human bodies & wind

😎Review https://bit.ly/435b42F
😎Paper arxiv.org/pdf/2305.10418.pdf
😎Project mmlab-ntu.github.io/project/layersnet

🎫 100% Mask-Free VIS 🎫

👉ETH Z unveils MaskFreeVIS: novel high-performing VIS without any mask annotations.

😎Review https://bit.ly/3Wg7CQB
😎Paper arxiv.org/pdf/2303.15904.pdf
😎Project www.vis.xyz/pub/maskfreevis/
😎Code github.com/SysCV/maskfreevis

🀄 Drag-GAN: user-friendly image-manipulation 🀄

👉 Manual deforming of (real and generated) images over pose, shape, expression and layout.

😎Review https://bit.ly/3BFyXlR
😎Paper arxiv.org/pdf/2305.10973.pdf
😎Project vcai.mpi-inf.mpg.de/projects/DragGAN
😎Code github.com/XingangPan/DragGAN

🗺️ AI-generated stereotypical men 🗺️

👉A thread about generating stereotypical person from 15 countries all around the world. And yes, Italian love Pizza.

😎 More https://bit.ly/3oo0t4c

🍶 AVOS Multiscale Encoder-Decoder ViT 🍶

👉 MED-VT, world's first Multiscale Encoder Decoder Video Transformer for AVOS

😎Review https://bit.ly/3MohFi1
😎Paper arxiv.org/pdf/2304.05930.pdf
😎Project rkyuca.github.io/medvt
😎Code github.com/rkyuca/medvt

🌊 Neural Dynamic Image-Based Rendering 🌊

👉 DynIBaR: synthesizing novel views from monocular video depicting a complex dynamic scene.

😎Review https://t.ly/90Kw
😎Paper arxiv.org/pdf/2211.11082.pdf
😎Project https://dynibar.github.io/
😎Code github.com/google/dynibar

🦁 Open Semantic Segmentation 🦁

👉SSSegmentation: open source supervised semantic segmentation toolbox based on #PyTorch

😎Review https://t.ly/ZE9q
😎Paper arxiv.org/pdf/2305.17091.pdf
😎Code github.com/SegmentationBLWX/sssegmentation

🎗️ 4D Humans with Transformers 🎗️

👉Novel approach to reconstruct and track humans (even in unusual poses)

😎Review https://t.ly/XGv_
😎Paper arxiv.org/pdf/2305.20091.pdf
😎Project shubham-goel.github.io/4dhumans/#
😎Code github.com/shubham-goel/4D-Humans

🗽 Neuralangelo Digital Twins. INSANE🗽

👉 A novel framework from #Nvidia for Hi-Fi 3D Digital twins.

😎Review https://t.ly/rxoF4
😎Project research.nvidia.com/labs/dir/neuralangelo
😎Paper research.nvidia.com/labs/dir/neuralangelo/paper.pdf

🦜 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

🌻 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

🏸 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