Detecting obstacles is crucial for safe and efficient autonomous driving. To this end, we present NVRadarNet, a deep neural network (DNN) that detects dynamic obstacles and drivable free space using automotive RADAR sensors. The network utilizes temporally accumulated data from multiple RADAR sensors to detect dynamic obstacles and compute their orientation in a top-down bird's-eye view (BEV). The network also regresses drivable free space to detect unclassified obstacles. Our DNN is the first of its kind to utilize sparse RADAR signals in order to perform obstacle and free space detection in real time from RADAR data only. The network has been successfully used for perception on our autonomous vehicles in real self-driving scenarios. The network runs faster than real time on an embedded GPU and shows good generalization across geographic regions.11Video at https://youtu.be/WlwJJMltoJY.

Autonomous driving requires the inference of actionable information such as detecting and classifying objects, and determining the drivable space. To this end, we present Multi-View LidarNet (MVLidarNet), a two-stage deep neural network for multi-class object detection and drivable space segmentation using multiple views of a single LiDAR point cloud. The first stage processes the point cloud projected onto a perspective view in order to semantically segment the scene. The second stage then processes the point cloud (along with semantic labels from the first stage) projected onto a bird's eye view, to detect and classify objects. Both stages use an encoder-decoder architecture. We show that our multi-view, multi-stage, multi-class approach is able to detect and classify objects while simultaneously determining the drivable space using a single LiDAR scan as input, in challenging scenes with more than one hundred vehicles and pedestrians at a time. The system operates efficiently at 150 fps on an embedded GPU designed for a self-driving car, including a postprocessing step to maintain identities over time. We show results on both KITTI and a much larger internal dataset, thus demonstrating the method's ability to scale by an order of magnitude.

The paper reports on the usage of deep learning methods for improving a Named Entity Recognition (NER) training corpus and for predicting and annotating new types in a test corpus. We show how the annotations in a type-based corpus of named entities (NE) were populated as occurrences within it, thus ensuring density of the training information. A deep learning model was adopted for discovering inconsistencies in the initial annotation and for learning new NE types. The evaluation results get improved after data curation, randomization and deduplication.

The paper continues investigations on the application of bidirectional echo state networks (BiESN) to the task of word sense disambiguation (WSD). Motivated by observations that the quality of the embedding vectors used to train the models influences to a significant degree their accuracy, here we propose the application of a single ESN reservoir to generate new potentially better embedding vectors with different dimensions. BiESN models for WSD of various reservoir sizes were trained using various combinations of new and original embeddings models for the input and/or output steps; the achieved accuracy is reported here. The results demonstrate increased WSD accuracy in several cases of newly derived embedding sets.

In this paper we present an approach for training verb subatom embeddings.For each verb we learn several embeddings rather than only one.These embeddings include the verb itself as well as embeddings for each grammatical role of this verb.To give an example, for the verb 'to give' we learn four embeddings: one for the lemma 'give', one for the subject, one for the direct object and one for the indirect object.We have exploited these grammatical role embeddings in order to add new syntagmatic relations to WordNet.The evaluation of the new relations quality has been done extrinsically through the Knowledge-based Word Sense Disambiguation task.

In this paper we present an analysis of different semantic relations extracted from WordNet, Extended WordNet and Sem-Cor, with respect to their role in the task of knowledge-based word sense disambiguation.The experiments use the same algorithm and the same test sets, but different variants of the knowledge graph.The results show that different sets of relations have different impact on the results: positive or negative.The beneficial ones are discussed with respect to the combination of relations and with respect to the test set.The inclusion of inference has only a modest impact on accuracy, while the addition of syntactic relations produces stable improvement over the baselines.