Modern self-driving autonomy systems heavily rely on deep learning. As a consequence, their performance is influenced significantly by the quality and richness of the training data. Data collecting platforms can generate many hours of raw data in a daily basis, however, it is not feasible to label everything. It is thus of key importance to have a mechanism to identify "what to label". Active learning approaches identify examples to label, but their interestingness is tied to a fixed model performing a particular task. These assumptions are not valid in self-driving, where we have to solve a diverse set of tasks (i.e., perception, and motion forecasting) and our models evolve over time frequently. In this paper we introduce a novel approach and propose a new data selection method that exploits a diverse set of criteria that quantize interestingness of traffic scenes. Our experiments on a wide range of tasks and models show that the proposed curation pipeline is able to select datasets that lead to better generalization and higher performance.
In this paper, we tackle the problem of depth completion from RGBD data. Towards this goal, we design a simple yet effective neural network block that learns to extract joint 2D and 3D features. Specifically, the block consists of two domain-specific sub-networks that apply 2D convolution on image pixels and continuous convolution on 3D points, with their output features fused in image space. We build the depth completion network simply by stacking the proposed block, which has the advantage of learning hierarchical representations that are fully fused between 2D and 3D spaces at multiple levels. We demonstrate the effectiveness of our approach on the challenging KITTI depth completion benchmark and show that our approach outperforms the state-of-the-art.
In this paper we propose to exploit multiple related tasks for accurate multi-sensor 3D object detection. Towards this goal we present an end-to-end learnable architecture that reasons about 2D and 3D object detection as well as ground estimation and depth completion. Our experiments show that all these tasks are complementary and help the network learn better representations by fusing information at various levels. Importantly, our approach leads the KITTI benchmark on 2D, 3D and BEV object detection, while being real time.
In this paper we propose a novel deep neural network that is able to jointly reason about 3D detection, tracking and motion forecasting given data captured by a 3D sensor. By jointly reasoning about these tasks, our holistic approach is more robust to occlusion as well as sparse data at range. Our approach performs 3D convolutions across space and time over a bird's eye view representation of the 3D world, which is very efficient in terms of both memory and computation. Our experiments on a new very large scale dataset captured in several north american cities, show that we can outperform the state-of-the-art by a large margin. Importantly, by sharing computation we can perform all tasks in as little as 30 ms.
In this paper we show that High-Definition (HD) maps provide strong priors that can boost the performance and robustness of modern 3D object detectors. Towards this goal, we design a single stage detector that extracts geometric and semantic features from the HD maps. As maps might not be available everywhere, we also propose a map prediction module that estimates the map on the fly from raw LiDAR data. We conduct extensive experiments on KITTI as well as a large-scale 3D detection benchmark containing 1 million frames, and show that the proposed map-aware detector consistently outperforms the state-of-the-art in both mapped and un-mapped scenarios. Importantly the whole framework runs at 20 frames per second.
In this paper, we propose a novel 3D object detector that can exploit both LIDAR as well as cameras to perform very accurate localization. Towards this goal, we design an end-to-end learnable architecture that exploits continuous convolutions to fuse image and LIDAR feature maps at different levels of resolution. Our proposed continuous fusion layer encode both discrete-state image features as well as continuous geometric information. This enables us to design a novel, reliable and efficient end-to-end learnable 3D object detector based on multiple sensors. Our experimental evaluation on both KITTI as well as a large scale 3D object detection benchmark shows significant improvements over the state of the art.
Sensor simulation is a key component for testing the performance of self-driving vehicles and for data augmentation to better train perception systems. Typical approaches rely on artists to create both 3D assets and their animations to generate a new scenario. This, however, does not scale. In contrast, we propose to recover the shape and motion of pedestrians from sensor readings captured in the wild by a self-driving car driving around. Towards this goal, we formulate the problem as energy minimization in a deep structured model that exploits human shape priors, reprojection consistency with 2D poses extracted from images, and a ray-caster that encourages the reconstructed mesh to agree with the LiDAR readings. Importantly, we do not require any ground-truth 3D scans or 3D pose annotations. We then incorporate the reconstructed pedestrian assets bank in a realistic LiDAR simulation system by performing motion retargeting, and show that the simulated LiDAR data can be used to significantly reduce the amount of annotated real-world data required for visual perception tasks.
In this paper, we propose an end-to-end self-driving network featuring a sparse attention module that learns to automatically attend to important regions of the input. The attention module specifically targets motion planning, whereas prior literature only applied attention in perception tasks. Learning an attention mask directly targeted for motion planning significantly improves the planner safety by performing more focused computation. Furthermore, visualizing the attention improves interpretability of end-to-end self-driving.
Feature selection is generally used as one of the most important pre-processing techniques in machine learning, as it helps to reduce the dimensionality of data and assists researchers and practitioners in understanding data. Thereby, better performance and reduced computational consumption, memory complexity and even data amount can be expected by utilizing feature selection. However, only few studies leverage the power of deep neural networks to solve the problem of feature selection. In this paper, we propose a feature mask module (FM-module) for feature selection based on a novel batch-wise attenuation and feature mask normalization. The proposed method is almost free from hyperparameters and can be easily integrated into common neural networks as an embedded feature selection method. Experiments on popular image, text and speech datasets have been shown that our approach is easy to use and has superior performance in comparison with other state-of-the-art deep learning based feature selection methods.
Recent years have seen a surge in the number of available frameworks for speech enhancement (SE) and recognition. Whether model-based or constructed via deep learning, these frameworks often rely in isolation on either time-domain signals or time-frequency (TF) representations of speech data. In this study, we investigate the advantages of each set of approaches by separately examining their impact on speech intelligibility and quality. Furthermore, we combine the fragmented benefits of time-domain and TF speech representations by introducing two new cross-domain SE frameworks. A quantitative comparative analysis against recent model-based and deep learning SE approaches is performed to illustrate the merit of the proposed frameworks.