Learning socially-aware motion representations is at the core of recent advances in human trajectory forecasting and robot navigation in crowded spaces. Yet existing methods often struggle to generalize to challenging scenarios and even output unacceptable solutions (e.g., collisions). In this work, we propose to address this issue via contrastive learning. Concretely, we introduce a social contrastive loss that encourages the encoded motion representation to preserve sufficient information for distinguishing a positive future event from a set of negative ones. We explicitly draw these negative samples based on our domain knowledge about socially unfavorable scenarios in the multi-agent context. Experimental results show that the proposed method consistently boosts the performance of previous trajectory forecasting, behavioral cloning, and reinforcement learning algorithms in various settings. Our method makes little assumptions about neural architecture designs, and hence can be used as a generic way to incorporate negative data augmentation into motion representation learning.
Pedestrians are arguably one of the most safety-critical road users to consider for autonomous vehicles in urban areas. In this paper, we address the problem of jointly detecting pedestrians and recognizing 32 pedestrian attributes. These encompass visual appearance and behavior, and also include the forecasting of road crossing, which is a main safety concern. For this, we introduce a Multi-Task Learning (MTL) model relying on a composite field framework, which achieves both goals in an efficient way. Each field spatially locates pedestrian instances and aggregates attribute predictions over them. This formulation naturally leverages spatial context, making it well suited to low resolution scenarios such as autonomous driving. By increasing the number of attributes jointly learned, we highlight an issue related to the scales of gradients, which arises in MTL with numerous tasks. We solve it by normalizing the gradients coming from different objective functions when they join at the fork in the network architecture during the backward pass, referred to as fork-normalization. Experimental validation is performed on JAAD, a dataset providing numerous attributes for pedestrian analysis from autonomous vehicles, and shows competitive detection and attribute recognition results, as well as a more stable MTL training.
In order to be globally deployed, autonomous cars must guarantee the safety of pedestrians. This is the reason why forecasting pedestrians' intentions sufficiently in advance is one of the most critical and challenging tasks for autonomous vehicles. This work tries to solve this problem by jointly predicting the intention and visual states of pedestrians. In terms of visual states, whereas previous work focused on x-y coordinates, we will also predict the size and indeed the whole bounding box of the pedestrian. The method is a recurrent neural network in a multi-task learning approach. It has one head that predicts the intention of the pedestrian for each one of its future position and another one predicting the visual states of the pedestrian. Experiments on the JAAD dataset show the superiority of the performance of our method compared to previous works for intention prediction. Also, although its simple architecture (more than 2 times faster), the performance of the bounding box prediction is comparable to the ones yielded by much more complex architectures. Our code is available online.
Drones or UAVs, equipped with different sensors, have been deployed in many places especially for urban traffic monitoring or last-mile delivery. It provides the ability to control the different aspects of traffic given real-time obeservations, an important pillar for the future of transportation and smart cities. With the increasing use of such machines, many previous state-of-the-art object detectors, who have achieved high performance on front facing cameras, are being used on UAV datasets. When applied to high-resolution aerial images captured from such datasets, they fail to generalize to the wide range of objects' scales. In order to address this limitation, we propose an object detection method called Butterfly Detector that is tailored to detect objects in aerial images. We extend the concept of fields and introduce butterfly fields, a type of composite field that describes the spatial information of output features as well as the scale of the detected object. To overcome occlusion and viewing angle variations that can hinder the localization process, we employ a voting mechanism between related butterfly vectors pointing to the object center. We evaluate our Butterfly Detector on two publicly available UAV datasets (UAVDT and VisDrone2019) and show that it outperforms previous state-of-the-art methods while remaining real-time.
Perceiving humans in the context of Intelligent Transportation Systems (ITS) often relies on multiple cameras or expensive LiDAR sensors. In this work, we present a new cost-effective vision-based method that perceives humans' locations in 3D and their body orientation from a single image. We address the challenges related to the ill-posed monocular 3D tasks by proposing a deep learning method that predicts confidence intervals in contrast to point estimates. Our neural network architecture estimates humans 3D body locations and their orientation with a measure of uncertainty. Our vision-based system (i) is privacy-safe, (ii) works with any fixed or moving cameras, and (iii) does not rely on ground plane estimation. We demonstrate the performance of our method with respect to three applications: locating humans in 3D, detecting social interactions, and verifying the compliance of recent safety measures due to the COVID-19 outbreak. Indeed, we show that we can rethink the concept of "social distancing" as a form of social interaction in contrast to a simple location-based rule. We publicly share the source code towards an open science mission.
Monocular and stereo vision are cost-effective solutions for 3D human localization in the context of self-driving cars or social robots. However, they are usually developed independently and have their respective strengths and limitations. We propose a novel unified learning framework that leverages the strengths of both monocular and stereo cues for 3D human localization. Our method jointly (i) associates humans in left-right images, (ii) deals with occluded and distant cases in stereo settings by relying on the robustness of monocular cues, and (iii) tackles the intrinsic ambiguity of monocular perspective projection by exploiting prior knowledge of human height distribution. We achieve state-of-the-art quantitative results for the 3D localization task on KITTI dataset and estimate confidence intervals that account for challenging instances. We show qualitative examples for the long tail challenges such as occluded, far-away, and children instances.
Scalable Vector Graphics (SVG) are ubiquitous in modern 2D interfaces due to their ability to scale to different resolutions. However, despite the success of deep learning-based models applied to rasterized images, the problem of vector graphics representation learning and generation remains largely unexplored. In this work, we propose a novel hierarchical generative network, called DeepSVG, for complex SVG icons generation and interpolation. Our architecture effectively disentangles high-level shapes from the low-level commands that encode the shape itself. The network directly predicts a set of shapes in a non-autoregressive fashion. We introduce the task of complex SVG icons generation by releasing a new large-scale dataset along with an open-source library for SVG manipulation. We demonstrate that our network learns to accurately reconstruct diverse vector graphics, and can serve as a powerful animation tool by performing interpolations and other latent space operations. Our code is available at https://github.com/alexandre01/deepsvg.
Since the past few decades, human trajectory forecasting has been a field of active research owing to its numerous real-world applications: evacuation situation analysis, traffic operations, deployment of social robots in crowded environments, to name a few. In this work, we cast the problem of human trajectory forecasting as learning a representation of human social interactions. Early works handcrafted this representation based on domain knowledge. However, social interactions in crowded environments are not only diverse but often subtle. Recently, deep learning methods have outperformed their handcrafted counterparts, as they learned about human-human interactions in a more generic data-driven fashion. In this work, we present an in-depth analysis of existing deep learning based methods for modelling social interactions. Based on our analysis, we propose a simple yet powerful method for effectively capturing these social interactions. To objectively compare the performance of these interaction-based forecasting models, we develop a large scale interaction-centric benchmark TrajNet++, a significant yet missing component in the field of human trajectory forecasting. We propose novel performance metrics that evaluate the ability of a model to output socially acceptable trajectories. Experiments on TrajNet++ validate the need for our proposed metrics, and our method outperforms competitive baselines on both real-world and synthetic datasets.
Semantic segmentation algorithms that can robustly segment objects across multiple camera viewpoints are crucial for assuring navigation and safety in emerging applications such as autonomous driving. Existing algorithms treat each image in isolation, but autonomous vehicles often revisit the same locations or maintain information from the immediate past. We propose a simple yet effective method for leveraging these image priors to improve semantic segmentation of images from sequential driving datasets. We examine several methods to fuse these temporal scene priors, and introduce a prior fusion network that is able to learn how to transfer this information. The prior fusion model improves the accuracy over the non-prior baseline from 69.1% to 73.3% for dynamic classes, and from 88.2% to 89.1% for static classes. Compared to models such as FCN-8, our prior method achieves the same accuracy with 5 times fewer parameters. We used a simple encoder decoder backbone, but this general prior fusion method could be applied to more complex semantic segmentation backbones. We also discuss how structured representations of scenes in the form of a scene graph could be leveraged as priors to further improve scene understanding.