🤯NVIDIA StyleGAN3 is out: source code available!🤯

👉Resolution and quality of images produced by GAN have seen rapid improvement recently. Despite their... 13 comments on LinkedIn

😵 New SOTA in video object segmentation is insane 😵

👉VOS aims at segment target instances in a video sequence. A high-performing #AI should be able... 127 comments on LinkedIn

🤪 The new SOTA in object tracking is awesome 🤪

👉Tracking by detection is the most effective paradigm for multi-object tracking in current. State-of-the-art... 11 comments on LinkedIn

😵😵𝑩𝒂𝒄𝒌 𝒕𝒐 𝒕𝒉𝒆 𝑭𝑬𝑨𝑻𝑼𝑹𝑬 with neural PixLoc. Source code available!😵😵

👉ETH Zurich & #Microsoft open-sourcing PixLoc, a neural network...

🧃🧃 A novel #AI paradigm for Video Instance Segmentation: Propose-Reduce just open-sourced 🧃🧃

👉Video instance segmentation (VIS) is a task to segment...

💎💎CMU is open-sourcing NeRS, a novel neural shape representation for surfaces💎💎

👉Recent history has seen the growth of implicit representations of...

🦑🦑 Non-Deep (∼10 layers) Neural Networks by Intel Labs are super-competitive 🦑🦑

👉Deep Neural Networks (DNNs) have revolutionized the fields of #AI...

👟👟❌❌ THIS SNEAKER DOES NOT EXIST ❌❌ 👟👟

Highlights:
✅Thousands of fake shoes via neural generation
✅Trained via #Nvidia StyleGAN2-ADA architecture
...

👽👽 #Nvidia just open-sourced DriveGAN, the most advanced neural simulator for #selfdrivingcars 👽👽

👉Realistic simulators are critical for training and… | 17 comments on LinkedIn

👱‍♀️👩‍🦰 上海交通大学 + Huawei open-sourcing CIPS-3D, a neural generator for 3D-aware synthetic faces 👱‍♀️👩‍🦰

👉StyleGAN achieved SOTA for generating HQ…

🔥🎽 #Fashion gets high with fooling #AI patches 🎽🔥

👉UNLABELED develops adversarial patterns to protect people from labeling and being…

🌳🛸University of Zurich open-sourced SVO Pro, one of the most solid framework for #slam and odometry 🛸🌳

👉SVO Pro features the support of different camera…

👶👧 The 𝘤𝘩𝘪𝘭𝘥-𝘱𝘢𝘳𝘦𝘯𝘵 paradigm in #AI to reach for the SOTA with simple fine-tuning and inversion 👶👧

👉Transfer Learning refers to the process in…

🔥🔥CLOOB outperforms OpenAI's CLIP🔥🔥

👉Contrastive Language-Image Pre-training (CLIP) yielded very impressive results at zero-shot transfer learning...

🎃🎃 (Dystopic) #AR Human Duplicator: #halloween is coming 🎃🎃

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Snap's #3D body tracking and reconstruction
✅"Hacking" a way to...

🤯🤯Rutgers University open-sourced SLATE, a novel neural architecture for zero-shot image generation without text🤯🤯

👉DALL·E shown an impressive ability...

👽👽#Nvidia open-sourced AugMax, a novel neural framework for data augmentation👽👽

👉Out-of-distribution samples present a challenge when deploying #...

🤯🤯Facebook #AI open-sourced GHNs: hypernetwork to predict the parameters for new neural networks in <1 second!🤯🤯

👉Is it possible to directly predict... 17 comments on LinkedIn

⏰⏰ German teams open-sourced Projected-GAN: training SOTA GANs 40x faster ⏰⏰

👉Generative Adversarial Networks produce high-quality images but are challenging... 10 comments on LinkedIn

🚙🚙Detailed report by #Valeo on adversarial attacks in #selfdrivingcars🚙🚙

👉Autonomous cars are expected to significantly reduce accidents. Despite...