Xavier Bourry

Flux inpaint, nicely done. Details here: https://lnkd.in/eTN43bKA

7

🧤🧤Transformer-based Interactive 4D Hands🧤🧤 👉4DHands is a novel and robust approach to recovering interactive hand meshes and their relative movement from monocular inputs. Authors: Beijing Normal University, Tsinghua & #Lenovo. No code announced yet 😢 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬: ✅RAT: relation-aware Two-Hand Tokenization  ✅Embedding positional relation info in tokenization ✅SIR: spatio-temporal Interaction Reasoning ✅RAT+SIR enhances stability of movements #artificialintelligence #machinelearning #ml #AI #deeplearning #computervision #AIwithPapers #metaverse 👉Discussion https://lnkd.in/dMgakzWm 👉Paper https://lnkd.in/dbUQcDkQ 👉Project https://4dhands.github.io/

209

3 Comments

It is lifecycle of any thing which came to existence on earth evolves, as only thing is constant is change. ofcourse as day by day XR Application and Experiences has started contributing to multiple streamlines and seems to be much better solution than any other possible alternatives to that experience, it is just that people have starting adapting to it, so without a doubt we can say that we are going to see another revolutionary boom very soon in the industry. #ar #vr #xr #mr #virtualreality #future #analysis #revolution

2

1 Comment

I have said for a while now that user data for passthru video and eye/gaze tracking in XR can't be fully exposed to most developers or amateurs. There is too much potential for misuse of biometric information. Biometrics need to be platform level that is only carefully exposed to verified developers. Otherwise users are going to get keylogged and datamined while using XR. ----- Wired: "Apple Vision Pro’s Eye Tracking Exposed What People Type The Vision Pro uses 3D avatars on calls and for streaming. These researchers used eye tracking to work out the passwords and PINs people typed with their avatars." "You can tell a lot about someone from their eyes. They can indicate how tired you are, the type of mood you’re in, and potentially provide clues about health problems. But your eyes could also leak more secretive information: your passwords, PINs, and messages you type. Today, a group of six computer scientists are revealing a new attack against Apple’s Vision Pro mixed reality headset where exposed eye-tracking data allowed them to decipher what people entered on the device’s virtual keyboard. The attack, dubbed GAZEploit and shared exclusively with WIRED, allowed the researchers to successfully reconstruct passwords, PINs, and messages people typed with their eyes. “Based on the direction of the eye movement, the hacker can determine which key the victim is now typing,” says Hanqiu Wang, one of the leading researchers involved in the work. They identified the correct letters people typed in passwords 77 percent of the time within five guesses and 92 percent of the time in messages. To be clear, the researchers did not gain access to Apple’s headset to see what they were viewing. Instead, they worked out what people were typing by remotely analyzing the eye movements of a virtual avatar created by the Vision Pro. This avatar can be used in Zoom calls, Teams, Slack, Reddit, Tinder, Twitter, Skype, and FaceTime. The researchers alerted Apple to the vulnerability in April, and the company issued a patch to stop the potential for data to leak at the end of July. It is the first attack to exploit people’s “gaze” data in this way, the researchers say. The findings underline how people’s biometric data—information and measurements about your body—can expose sensitive information and be used as part of the burgeoning surveillance industry. “These technologies … can inadvertently expose critical facial biometrics, including eye-tracking data, through video calls where the user’s virtual avatar mirrors their eye movements,” the researchers write in a preprint paper detailing their findings. Wang says the work relies on two biometrics that can be extracted from recordings of a Persona: the eye aspect ratio (EAR) and eye gaze estimation. (the research was completed by Wang, Siqi Dai, Max Panoff, and Shuo Wang from Univ of Florida, Haoqi Shan and Zihao Zhan from Texas Tech.)" Wired: https://lnkd.in/gv6BaHDD #xr #biometrics #apple #avp #visionpro

11

3 Comments

🪩🪩MimicMotion: HQ Motion Generation🪩🪩 👉#Tencent opens a novel controllable video generation framework, dubbed MimicMotion, which can generate HQ videos of arbitrary length mimicking specific motion guidance. Source Code available💙 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬: ✅MimicMotion: I2V diffusion + confidence-aware pose ✅Rich details, good temporal smoothness & long video ✅Better pose by novel confidence-aware strategy ✅Correcting erroneous pose guidance during inference ✅Hand region enhancement to alleviate hand distortion ✅Position-aware progressive latent fusion #artificialintelligence #machinelearning #ml #AI #deeplearning #computervision #AIwithPapers #metaverse 👉Discussion https://lnkd.in/dMgakzWm 👉Paper arxiv.org/pdf/2406.19680 👉Project https://lnkd.in/eW-CMg_C 👉Code https://lnkd.in/eZ6SC2bc

231

Realistic 3D shapes from text prompts? MIT researchers cracked the code. Their simple fix to Score Distillation enables sharper, higher-quality 3D models without costly retraining or complex postprocessing. This could revolutionize virtual reality, filmmaking, and engineering design. Are you ready for the 3D AI revolution? #GenerativeAI #3DModeling #ComputerVision #AIInnovation #FutureOfDesign What 3D creations would you bring to life with this technology?

2

With Mesh Morpher`s latest version for UE 5.4 (Release) we can subdivide Meta Humans sections (materials) and preserve the Morph Targets as well. This allows us to further customize our Meta Humans the way we want. #ue5 #metahuman #meshmorpher

25

2 Comments

Absolutely exciting times! The French AI space keeps heating up and I’m thrilled at the innovation we keep seeing…so much more to more expect! Huge welldone to the entire Kyutai team! Loved watching this demo of your text-to-speech AI model! Yes to making our daily lives easier! ✨✨✨ Thanks for sharing Andreas Horn strategywithifeoluwa.com On eagles' wings #strategywithifeoluwa #swif #startups #ideation #strategy #implementation #launch #founders #aspiringfounders #professionals #value #customers #gotomarketstrategy #productmarketfit #execution

1

4 Comments

Want to understand how the Meta Orion and Snapchat spectacles prototypes of AR glasses compare ? Look at this excellent video from Marques Brownlee https://lnkd.in/eWEiWSNs Want to update your vocabulary when speaking of immersive devices and how they give you or not a sense of presence ? Read the comment I made on the video and copied below 👇 (Referring Mel Slater) —- Hi Marques, very interesting comparison, thank you You are wondering at minute 15 if you should use the word immersion to describe what happens to you when the lack of FOV cuts the overlay Let me point you to one of the most important scientific paper on the topic, a reference in the research community on VR (I’m part of it since almost 30 years): https://lnkd.in/eRGciSfz Besides immersion, the other useful words here are: presence, plausibility and illusion In short: - Immersion is a way to compare technical capacities of devices: the Meta glasses are more immersive in terms of FOV, the snap glasses are more immersive in terms of resolution … - Presence is what it makes to the user in terms of feeling, in its mind Presence is made of 2 factors: place illusion and plausibility illusion - Place illusion is the perceptual (scientists say “sensory-motor”) part: it is the sentiment of non mediation of the device, or the fact that you believe things are here - Plausibility illusion is more cognitive: it has to do with storytelling, consistency of the experience, absence of gliches over time. It depends more on the content of the experience and less on the container. Place illusion is easy to recover when it’s broken: when your overlay is being cut on the side it breaks the presence, but when recentered its presence is restored On the other hand, plausibility illusion is hard to restore when broken. Hope this helps ! Cheers David Nahon. @iVEvangelist

8

1 Comment

🏗️🏗️ #Adobe Diffusive Instant TurboEdit 🏗️🏗️ 👉Adobe unveils a novel real-time text-based disentangled real image editing method built upon 4-step SDXL Turbo. SOTA HQ image editing using ultra fast few-step diffusion. No code announced but easy to guess it will be released in commercial tools. 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬: ✅Disentangled controls in few-step diffusion ✅Hyper-realistic text-guided editing in real-time ✅Only 8 NFEs in inversion (ot-cost), 4 per edit ✅Outperforming SOTA multi-step diffusion editing #AI #AIwithPapers #artificialintelligence #machinelearning #ml #deeplearning #computervision #metaverse 👉Discussion https://lnkd.in/dMgakzWm 👉Paper https://lnkd.in/dVs9RcCK 👉Project https://lnkd.in/dGCqwh9Z 👉Code 😢

247

7 Comments

Has anyone used Meta Quest 3 or other AR/VR/mixed reality headsets? I'm looking for something that could be useful as both a capture device (sensors) and for AR/VR/MR/Vision related projects. If you have recommendations, please leave a comment. #AR #VR #ComputerVision #mixedreality

2

2 Comments

🎢🎢SoloParkour: robo-quadruped parkour🎢🎢 👉LAAS-CNRS unveils a novel RL approach to perform agile skills that are reminiscent of parkour, such as walking, climbing high steps, leaping over gaps, and crawling under obstacles. Data and Code available💙 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬: ✅Novel method for training end-to-end visual policies ✅Goal: achieving agile and safe quadruped locomotion ✅Robo-parkour as a constrained novel RL problem ✅Skills such as walking, climbing, leaping & crawling #artificialintelligence #machinelearning #ml #AI #deeplearning #computervision #AIwithPapers #metaverse 👉Discussion https://lnkd.in/dMgakzWm 👉Paper arxiv.org/pdf/2409.13678 👉Project https://lnkd.in/dpyKf2SE 👉Code https://lnkd.in/dy9mqJKn

106

3 Comments

Fluid and seamless 2D shapes, instant feedback, and responsive interactions— Rive is redefining what's possible in animation. Perfect for crafting custom, high-end UI interfaces, especially in HMI and industrial applications. 🖥️🤖 We’re truly excited to bring this powerful tool into more of our projects. The possibilities are endless! 🔧💡 #UIUX #Animation #HMI #IndustrialDesign #Rive #DigitalInnovation #DesignTools #CreativeTech

16

1 Comment

Hot off the AWE stage, this announcement from Niantic, Inc. & Scaniverse is for the 3D devs and reality capture fans: Introducing SPZ, the open source format for 3D Gaussian Splatting that shrinks file sizes up to 90% compared to PLY. For those not in 3D or game dev - this means high-resolution, immersive 3D scans that can be used in AR games and spatial computing apps just got a lot smaller, simpler to use, faster to load, and better for sharing. Better assets for better games and experiences! #3DGS #3DGaussianSplatting #3DGaussianSplats #3DGaussianSplat #Splat #ar #vr #xr #spatialcomputing #augmentedreality #virtualreality #AWE2024 #dev #3dDev #gameDev #RealityCapture

14

4 Comments

💡💡DifFRelight: Diffusion Models Relighting💡💡 👉#Netflix (Eyeline Studios) unveils DifFRelight, a novel framework for free-viewpoint facial performance relighting via diffusion model. Precise lighting control, enabling high-fidelity relit facial images from flat-lit inputs. No code announced but amazing results. 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬: ✅Free-viewpoint face relighting via diffusion ✅Spatially-aligned conditioning of flat-lit & noise ✅Unified control by integrating novel area lighting ✅HDRI composition via multiple directional lights #artificialintelligence #machinelearning #ml #AI #deeplearning #computervision #AIwithPapers #metaverse 👉Discussion https://lnkd.in/dMgakzWm 👉Paper arxiv.org/pdf/2410.08188 👉Project https://lnkd.in/dku5crPY

179

5 Comments

💥 Towards Modular LLMs by Building and Reusing a Library of LoRAs Oleksiy Ostapenko, Zhan Su, Edoardo Maria Ponti, Laurent Charlin, Nicolas Le Roux, Matheus Pereira, Lucas Caccia, Alessandro Sordoni Abstract The growing number of parameter-efficient adaptations of a base large language model (LLM) calls for studying whether we can reuse such trained adapters to improve performance for new tasks. We study how to best build a library of adapters given multi-task data and devise techniques for both zero-shot and supervised task generalization through routing in such library. We benchmark existing approaches to build this library and introduce model-based clustering, MBC, a method that groups tasks based on the similarity of their adapter parameters, indirectly optimizing for transfer across the multi-task dataset. To re-use the library, we present a novel zero-shot routing mechanism, Arrow, which enables dynamic selection of the most relevant adapters for new inputs without the need for retraining. We experiment with several LLMs, such as Phi-2 and Mistral, on a wide array of held-out tasks, verifying that MBC-based adapters and Arrow routing lead to superior generalization to new tasks. We make steps towards creating modular, adaptable LLMs that can match or outperform traditional joint training. 👉 https://lnkd.in/daFqETww #machinelearning

2

#OnChain almost always implies open source open source works best on open money chain of custody of training data on chain and chain of custody of outputs on chain #OnChainAi #AiOnChain #ChainedAI

1