Representation is a key theme in deep learning—and machine learning more broadly—applied to geometry. Assorted means of communicating a shape to and from a deep network present varying tradeoffs between efficiency, quality, and applicability. Whil...

We explore a specific skill in Visual Question Answering that is important for applications involving aiding visually impaired users – answering questions about everyday images that involve reading and reasoning about text in these images. We find...

Datasets We validate our approach on different datasets and tasks. We focus on dense prediction tasks that can be approached with fully convolutional architectures. Most of the experiments are carried out on the PASCAL [Eve+12] benchmark, which is...

SpecAugment greatly improves the performance of ASR networks. We are able to obtain state-of-the-art results on the LibriSpeech 960h and Switchboard 300h tasks on end-to-end LAS networks by augmenting the training set using simple hand-crafted pol...

We introduced the problem of detecting and predicting injuries in pitchers from only video data. However, there are many possible limitations and extensions to our work. While we showed that CNN can reliably detect and predicted injuries, due to t...

We present RL-VAE, a graph-to-graph variational autoencoder that uses reinforcement learning to decode molecular graphs from latent embeddings. Methods have been described previously for graph-to-graph autoencoding, but these approaches require so...

In this paper we develop a novel deep structured model for 3D scene flow estimation. We focus on the self-driving scenario where the motion of the scene can be composed by estimating the 3D rigid motion of each actor. We first exploit deep learnin...

In this paper, we brought up to the literature the discussion about the Elsagate phenomenon, and we proposed the first solution to solve the problem. Inspired by the most recent approaches in sensitive content detection, we evaluated popular DLAs ...

We have shown that diverse sources of both laboratory and astronomical information about cold, dense, catalyzed matter can be incorporated flexibly within a straightforward, rigorous, and practical Bayesian framework. We treat carefully the constr...

We propose CornerNet-Lite which is a combination of two efficient variant of CornerNet: CornerNet-Saccade and CornerNet-Squeeze. Together these contributions for the first time reveal the potential of keypoint-based detection to be useful for appl...

In this work, we propose KPConv, a convolution that operates on point clouds. KPConv takes radius neighborhoods as input and processes them with weights spatially located by a small set of kernel points. We define a deformable version of this conv...

The recent boom in creating machine learnable fingerprints for atomistic systems, or descriptors, has led to a plethora of available options for materials science. The software implementations for these descriptors is, however, often scattered acr...

In this paper, after reviewing recent learning-based works on cell counting and detection, the common step of coding raw dot labels is extracted and discussed. Two center coding criteria are proposed: entropy and reversibility. These two criteria ...

In summary, given the inter-user variability of force applied over the surface of the robot, the personalization of the system response based on the continuous input of the user would allow to account, not only for the particularities of each inte...

Variational Autoencoders(VAEs) have already achieved great results on image generation and recently made promising progress on music generation. However, the generation process is still quite difficult to control in the sense that the learned late...

An analysis of the literature shows that there are two types of non-memristive models that have been widely used in the modeling of so-called "memristive" neural networks. Here, we demonstrate that such models have nothing in common with the conce...

The authors are grateful to an anonymous referee and the boarder member for the constructive comments, which have significantly improved the presentation of the paper. The work of Y. Xu was partially supported by National Natural Science Foundatio...

In this paper, we propose a recursive and SQL clause-wise decoding neural architecture for the complex and cross-domain text-to-SQL generation. We evaluate our model with Spider dataset, and the experimental result shows that our model significant...

This study proposed a passivity-based nonlinear admittance controller that fully captures the nonlinearities of robot dynamics. The proposed control structure can be represented as a feedback interconnection of passive subsystems, and therefore di...

We introduced the idea of applying attentional operations for deep learning models on lane graphs. Our results indicate that these particular NN architectures can be useful on the lane level, where first-principles models are difficult to derive.