⚗️ SemAbs: 3D Scene Understanding ⚗️

👉Framework that equips 2D Vision-Language Models (VLMs) with new 3D spatial capabilities

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅2D VLMs with 3D reasoning skills
✅ViTs Efficient MS Relevancy Extraction
✅Novel Open-World understanding tasks
✅Completing partially observed objects
✅Finding hidden objects from language

More: https://bit.ly/3PYYk7d

🦚 TinyCD: Neural Change Detection 🦚

👉TinyCD: new SOTA in change detection with up to 150x fewer parameters.

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅SOTA with up to 150X fewer params
✅Mixing blocks for s.t. cross-correlation
✅PW-MLP for pixel wise classification
✅MAMB: novel block for skip connection

More: https://bit.ly/3zFEngk

🦊 3D-Aware "StyleGANv2" version 🦊

👉Upgrading StyleGANv2 into a novel 3D-aware GAN with just a minimal set of changes🤯

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅MPI-like 3D-aware GAN w/ single-view
✅GMPI: generative multiplane image
✅2D GAN 3D-aware with a minimal changes
✅Encoding 3D-aware inductive biases

More: https://bit.ly/3OJ5gnS

📺 NeRF-ing "The Big Bang Theory" 📺

👉Berkeley unveils an approach for accurate estimation of actor’s 3D pose & location

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Input: images across the whole season
✅3D context (i.e. cams, structure, body)
✅Integrating context in 3D estimation
✅Re-ID, gaze, cinematography, pic editing
✅Knock, Knock, Penny!

More: https://bit.ly/3OLuaUb

🎩ShAPO: SOTA in object understanding🎩

👉Joint multi-object detection, #3D texture, 6D object pose & size estimation.

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Disentangled shape & appearance
✅Efficient octree-based differentiable
✅Object-centric understanding pipeline
✅Detection, reconstruction , 6D & size
✅SOTA in reconstruction & pose est.

More: https://bit.ly/3oHN5EQ

🏙️ CityNeRF: Neural Rendering of City Scenes 🏙️

👉Progressive NeRF model and training set on city-scenes

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅BungeeNeRF: novel progressive NeRF
✅Details on drastically varied scales
✅Growing with residual block structure
✅Inclusive multi-level data supervision

More: https://bit.ly/3cS9vk7

🍦🍦 Rewriting Geometry of GAN 🍦🍦

👉Drive GAN synthesizing many unseen objects with the desired shape

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅User-friendly "warping" with geometry
✅Low-rank update to layer for editing
✅Latent augmentation based on style-mix
✅Endless objects with defined changes
✅Latent space interpolation, image editing

More: https://bit.ly/3zIfOj8

🍏🍏 GAUDI: the Neural Architect 🍏🍏

👉Novel generative model for immersive 3D scenes from a moving camera

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Hundreds of thousands pics/scenes
✅Novel denoising optimization objective
✅New SOTA across multiple datasets
✅Un/conditional on images/text

More: https://bit.ly/3Bt65ye

🚜NeDDF: the NeRF evolution!🚜

👉Novel 3D representation that reciprocally constrains distance & density fields

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅NeRF provides no distance
✅Extending for arbitrary density
✅Density via dist-field & gradient
✅Alleviating the instability

More: https://bit.ly/3Bte8LC

🔥AND/OR: Composable Diffusion Models🔥

👉Novel neural compositional generation via Composable Diffusion Models

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅DM as energy-based models
✅Connecting diffusion models
✅Conjunction & negation, on top of DM
✅Zero-shot combinatorial generalization

More: https://bit.ly/3PYv1Cs

🔥 MobileNeRF is out -> Pure Fire! 🔥

👉MobileNeRF is out: the mobile evolution of NeRF via textured polygons.

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Same quality, 10x faster than SNeRG
✅Memory-- by storing surface textures
✅Integrated GPUs: less memory/power
✅Suitable for browser & viewer is HTML

More: https://bit.ly/3PUKPWy

🔥🔥UPDATE🔥🔥

Code Released: https://github.com/andreas128/RePaint

🧣NeRF for Outdoor Scene Relighting🧣

👉NeRF-OSR: the first neural radiance fields approach for outdoor scene relighting

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅NeRF-method for outdoor relighting
✅Simultaneous illumination/viewpoint
✅Control over shading, shadow, albedo
✅Self-Supervised training from outdoor
✅Dataset: 3240 viewpoints, 110+ times

More: https://bit.ly/3vBiH2G

👩‍🦰 Real-Time Neural Hair 👩‍🦰

👉Accurate hair geometry & appearance from multi-pics

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Bonn, CMU and Reality Labs
✅Photorealistic Real-Time render
✅HQ strand geometry/appearance
✅Novel scalp texture description
✅Intuitive manipulation of 3D hair

More: https://bit.ly/3vBiH2G

🚀 #VR by NASA - 1985 🚀

👉Q: is #VR the technology that developed least in the last 40 years? 🤔

Let's talk: https://bit.ly/3JxDZ7i

🔥 MinVIS, a new SOTA is out 🔥

👉#Nvidia miniVIS: no video-based architectures nor training procedures🤯

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Video architecture/train not required
✅MinVIS outperforms the previous SOTA
✅Occluded VIS (OVIS): >10% improvement
✅1% of labeled frames >> fully-supervised

More: https://bit.ly/3pcYzk1

🔥🔥MultiNeRF: three NeRFs are out!🔥🔥

👉Google opens the code of three #cvpr2022 papers: Mip-NeRF 360, Ref-NeRF, RawNeRF

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Paper_1: Mip-NeRF 360
✅Paper_2: Ref-NeRF
✅Paper_3: NeRF in the Dark

More: https://bit.ly/3QjpRRc

☀️LocoProp: Neural Layers Composition☀️

👉Google AI unveils LocoProp: novel neural paradigm for modular composition of layers.

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Backprop++ via Local Loss Optimization
✅Layer-based w-reg, target output, loss
✅Multiple local update via first-order opt.
✅Superior performance and efficiency

More: https://bit.ly/3Q40YJn

🔥PCVOS: clip-wise mask VOS🔥

👉PCVOS: new semi-supervised video object segmentation method

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Reformulating semi-supervised VOS
✅Novel per-clip inference perspective
✅Clip-wise operation on intra-clip
✅PCVOS: model for per-clip inference
✅New SOTA on multiple benchmarks

More: https://bit.ly/3vJtmbz

🍑 World-Object Detection via ViT 🍑

👉Google unveils OWL-ViT: open-vocabulary detector based on ViTs 🤯

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅ViTs for Open-World Localization
✅Img-level to open-vocabulary detection
✅SOTA one-shot (img.cond.) detection

More: https://bit.ly/3Sy3jOj

🎹🎹 Learning Piano in #AR 🎹🎹

👉PianoVision (on #META #Quest2) accelerates the piano learning via Passthrough #AR & hand tracking

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Sheet Insight to learn sight-read
✅MIDI keyboard connectivity
✅Air piano for no physical pianos
✅Multiplayer Music Instruction
✅PianoVision Music Hall in #VR

More: https://bit.ly/3zYvwGX