Novel categories names:['High_Jump', 'Front_Crawl', 'Pole_V ault', 'Hammer_Throw', All experiments are conducted under the 16-shot setting. An incremental bayesian approach tested on 101 object categories. Conditional prompt learning for vision-language models.

The paper introduces Field Probing Neural Networks, an extrinsic construction based on 3D volumetric fields that circumvents limitations of voxel based approaches. The paper is well written and I find the idea rather interesting, despite not having a huge gap in raw performance (but a huge one in terms of computational resources). There are many repetitions (mostly nouns) in the text which could be removed to make it easier to read. "However, existing 3D CNN pipelines" - I would remove However. Figure 1: An visualization - A visualization. I would like the authors to make clear that their construction is purely extrinsic and that therefore in case of deformable objects it will not be invariant to isometries.

We present a unified hard-constraint framework for solving geometrically complex PDEs with neural networks, where the most commonly used Dirichlet, Neumann, and Robin boundary conditions (BCs) are considered. Specifically, we first introduce the "extra fields'' from the mixed finite element method to reformulate the PDEs so as to equivalently transform the three types of BCs into linear forms. Based on the reformulation, we derive the general solutions of the BCs analytically, which are employed to construct an ansatz that automatically satisfies the BCs. With such a framework, we can train the neural networks without adding extra loss terms and thus efficiently handle geometrically complex PDEs, alleviating the unbalanced competition between the loss terms corresponding to the BCs and PDEs. We theoretically demonstrate that the "extra fields'' can stabilize the training process. Experimental results on real-world geometrically complex PDEs showcase the effectiveness of our method compared with state-of-the-art baselines.

This article describes how Microsoft wants to break Google's Dominance in the Search Engine Business.

High-fidelity reconstruction of fluids from sparse multiview RGB videos remains a formidable challenge due to the complexity of the underlying physics as well as complex occlusion and lighting in captures. Existing solutions either assume knowledge of obstacles and lighting, or only focus on simple fluid scenes without obstacles or complex lighting, and thus are unsuitable for real-world scenes with unknown lighting or arbitrary obstacles. We present the first method to reconstruct dynamic fluid by leveraging the governing physics (ie, Navier -Stokes equations) in an end-to-end optimization from sparse videos without taking lighting conditions, geometry information, or boundary conditions as input. We provide a continuous spatio-temporal scene representation using neural networks as the ansatz of density and velocity solution functions for fluids as well as the radiance field for static objects. With a hybrid architecture that separates static and dynamic contents, fluid interactions with static obstacles are reconstructed for the first time without additional geometry input or human labeling. By augmenting time-varying neural radiance fields with physics-informed deep learning, our method benefits from the supervision of images and physical priors. To achieve robust optimization from sparse views, we introduced a layer-by-layer growing strategy to progressively increase the network capacity. Using progressively growing models with a new regularization term, we manage to disentangle density-color ambiguity in radiance fields without overfitting. A pretrained density-to-velocity fluid model is leveraged in addition as the data prior to avoid suboptimal velocity which underestimates vorticity but trivially fulfills physical equations. Our method exhibits high-quality results with relaxed constraints and strong flexibility on a representative set of synthetic and real flow captures.

Seagrass meadows are a key ecosystem of the Great Barrier Reef World Heritage Area, providing one of the natural heritage attributes underpinning the reef’s outstanding universal value. We reviewed approaches employed to date to create maps of seagrass meadows in the optically complex waters of the Great Barrier Reef and explored enhanced mapping approaches with a focus on emerging technologies, and key considerations for future mapping. Our review showed that field-based mapping of seagrass has traditionally been the most common approach in the GBRWHA, with few attempts to adopt remote sensing approaches and emerging technologies. Using a series of case studies to harness the power of machine- and deep-learning, we mapped seagrass cover with PlanetScope and UAV-captured imagery in a variety of settings. Using a machine-learning pixel-based classification coupled with a bootstrapping process, we were able to significantly improve maps of seagrass, particularly in low cover, fragmented and complex habitats. We also used deep-learning models to derive enhanced maps from UAV imagery. Combined, these lessons and emerging technologies show that more accurate and efficient seagrass mapping approaches are possible, producing maps of higher confidence for users and enabling the upscaling of seagrass mapping into the future.

Inspired by recent successes of deep learning in computer vision, we propose a novel framework for encoding time series as different types of images, namely, Gramian Angular Summation/Difference Fields (GASF/GADF) and Markov Transition Fields (MTF). This enables the use of techniques from computer vision for time series classification and imputation. We used Tiled Convolutional Neural Networks (tiled CNNs) on 20 standard datasets to learn high-level features from the individual and compound GASF-GADF-MTF images. Our approaches achieve highly competitive results when compared to nine of the current best time series classification approaches. Inspired by the bijection property of GASF on 0/1 rescaled data, we train Denoised Auto-encoders (DA) on the GASF images of four standard and one synthesized compound dataset. The imputation MSE on test data is reduced by 12.18% – 48.02% when compared to using the raw data. An analysis of the features and weights learned via tiled CNNs and DAs explains why the approaches work.

Eight years ago a machine learning algorithm learned to identify a cat--and it stunned the world. A few years later AI could accurately translate languages and take down world champion Go players. Now, machine learning has begun to excel at complex multiplayer video games like Starcraft and Dota 2 and subtle games like poker. AI, it would appear, is improving fast. But how fast is fast, and what's driving the pace?

Technical quarrels between quantum computing experts rarely escape the field's rarified community. Late Monday, though, IBM's quantum team picked a highly public fight with Google. In a technical paper and blogpost, IBM took aim at potentially history-making scientific results accidentally leaked from a collaboration between Google and NASA last month. That draft paper claimed Google had reached a milestone dubbed "quantum supremacy"--a kind of drag race in which a quantum computer proves able to do something a conventional computer can't. Monday, Big Blue's quantum PhD's said Google's claim of quantum supremacy was flawed.

An article by Apple describing how the use machine learning to improve HomePod's signal processing system to remove echo, background noise and separate sound sources.