Vincent Frémont

Minh-Quan Dao, Julie Stephany Berrio, Vincent Frémont et al.

Occlusion is a major challenge for LiDAR-based object detection methods. This challenge becomes safety-critical in urban traffic where the ego vehicle must have reliable object detection to avoid collision while its field of view is severely reduced due to the obstruction posed by a large number of road users. Collaborative perception via Vehicle-to-Everything (V2X) communication, which leverages the diverse perspective thanks to the presence at multiple locations of connected agents to form a complete scene representation, is an appealing solution. State-of-the-art V2X methods resolve the performance-bandwidth tradeoff using a mid-collaboration approach where the Bird-Eye View images of point clouds are exchanged so that the bandwidth consumption is lower than communicating point clouds as in early collaboration, and the detection performance is higher than late collaboration, which fuses agents' output, thanks to a deeper interaction among connected agents. While achieving strong performance, the real-world deployment of most mid-collaboration approaches is hindered by their overly complicated architectures, involving learnable collaboration graphs and autoencoder-based compressor/ decompressor, and unrealistic assumptions about inter-agent synchronization. In this work, we devise a simple yet effective collaboration method that achieves a better bandwidth-performance tradeoff than prior state-of-the-art methods while minimizing changes made to the single-vehicle detection models and relaxing unrealistic assumptions on inter-agent synchronization. Experiments on the V2X-Sim dataset show that our collaboration method achieves 98\% of the performance of an early-collaboration method, while only consuming the equivalent bandwidth of a late-collaboration method.

Andrew Caunes, Thierry Chateau, Vincent Fremont

Semantic and panoptic occupancy prediction for road scene analysis provides a dense 3D representation of the ego vehicle's surroundings. Current camera-only approaches typically rely on costly dense 3D supervision or require training models on data from the target domain, limiting deployment in unseen environments. We propose FreeOcc, a training-free pipeline that leverages pretrained foundation models to recover both semantics and geometry from multi-view images. FreeOcc extracts per-view panoptic priors with a promptable foundation segmentation model and prompt-to-taxonomy rules, and reconstructs metric 3D points with a reconstruction foundation model. Depth- and confidence- aware filtering lifts reliable labels into 3D, which are fused over time and voxelized with a deterministic refinement stack. For panoptic occupancy, instances are recovered by fitting and merging robust current-view 3D box candidates, enabling instance-aware occupancy without any learned 3D model. On Occ3D-nuScenes, FreeOcc achieves 16.9 mIoU and 16.5 RayIoU train-free, on par with state-of-the-art weakly supervised methods. When employed as a pseudo-label generation pipeline for training downstream models, it achieves 21.1 RayIoU, surpassing the previous state-of-the-art weakly supervised baseline. Furthermore, FreeOcc sets new baselines for both train-free and weakly supervised panoptic occupancy prediction, achieving 3.1 RayPQ and 3.9 RayPQ, respectively. These results highlight foundation-model-driven perception as a practical route to training-free 3D scene understanding.

Minh-Quan Dao, Vincent Frémont, Elwan Héry

Low-resolution point clouds are challenging for object detection methods due to their sparsity. Densifying the present point cloud by concatenating it with its predecessors is a popular solution to this challenge. Such concatenation is possible thanks to the removal of ego vehicle motion using its odometry. This method is called Ego Motion Compensation (EMC). Thanks to the added points, EMC significantly improves the performance of single-frame detectors. However, it suffers from the shadow effect that manifests in dynamic objects' points scattering along their trajectories. This effect results in a misalignment between feature maps and objects' locations, thus limiting performance improvement to stationary and slow-moving objects only. Scene flow allows aligning point clouds in 3D space, thus naturally resolving the misalignment in feature spaces. By observing that scene flow computation shares several components with 3D object detection pipelines, we develop a plug-in module that enables single-frame detectors to compute scene flow to rectify their Bird-Eye View representation. Experiments on the NuScenes dataset show that our module leads to a significant increase (up to 16%) in the Average Precision of large vehicles, which interestingly demonstrates the most severe shadow effect. The code is published at https://github.com/quan-dao/pc-corrector.

Minh-Quan Dao, Elwan Héry, Vincent Frémont

Recent advances in 3D object detection are made by developing the refinement stage for voxel-based Region Proposal Networks (RPN) to better strike the balance between accuracy and efficiency. A popular approach among state-of-the-art frameworks is to divide proposals, or Regions of Interest (ROI), into grids and extract features for each grid location before synthesizing them to form ROI features. While achieving impressive performances, such an approach involves several hand-crafted components (e.g. grid sampling, set abstraction) which requires expert knowledge to be tuned correctly. This paper proposes a data-driven approach to ROI feature computing named APRO3D-Net which consists of a voxel-based RPN and a refinement stage made of Vector Attention. Unlike the original multi-head attention, Vector Attention assigns different weights to different channels within a point feature, thus being able to capture a more sophisticated relation between pooled points and ROI. Our method achieves a competitive performance of 84.85 AP for class Car at moderate difficulty on the validation set of KITTI and 47.03 mAP (average over 10 classes) on NuScenes while having the least parameters compared to closely related methods and attaining an inference speed at 15 FPS on NVIDIA V100 GPU. The code is released at https://github.com/quan-dao/APRO3D-Net.

Minh-Quan Dao, Holger Caesar, Julie Stephany Berrio et al.

Occlusion presents a significant challenge for safety-critical applications such as autonomous driving. Collaborative perception has recently attracted a large research interest thanks to the ability to enhance the perception of autonomous vehicles via deep information fusion with intelligent roadside units (RSU), thus minimizing the impact of occlusion. While significant advancement has been made, the data-hungry nature of these methods creates a major hurdle for their real-world deployment, particularly due to the need for annotated RSU data. Manually annotating the vast amount of RSU data required for training is prohibitively expensive, given the sheer number of intersections and the effort involved in annotating point clouds. We address this challenge by devising a label-efficient object detection method for RSU based on unsupervised object discovery. Our paper introduces two new modules: one for object discovery based on a spatial-temporal aggregation of point clouds, and another for refinement. Furthermore, we demonstrate that fine-tuning on a small portion of annotated data allows our object discovery models to narrow the performance gap with, or even surpass, fully supervised models. Extensive experiments are carried out in simulated and real-world datasets to evaluate our method.

Andrew Caunes, Thierry Chateau, Vincent Frémont

Semantic segmentation of 3D LiDAR point clouds, essential for autonomous driving and infrastructure management, is best achieved by supervised learning, which demands extensive annotated datasets and faces the problem of domain shifts. We introduce a new 3D semantic segmentation pipeline that leverages aligned scenes and state-of-the-art 2D segmentation methods, avoiding the need for direct 3D annotation or reliance on additional modalities such as camera images at inference time. Our approach generates 2D views from LiDAR scans colored by sensor intensity and applies 2D semantic segmentation to these views using a camera-domain pretrained model. The segmented 2D outputs are then back-projected onto the 3D points, with a simple voting-based estimator that merges the labels associated to each 3D point. Our main contribution is a global pipeline for 3D semantic segmentation requiring no prior 3D annotation and not other modality for inference, which can be used for pseudo-label generation. We conduct a thorough ablation study and demonstrate the potential of the generated pseudo-labels for the Unsupervised Domain Adaptation task.

Katie Z Luo, Minh-Quan Dao, Zhenzhen Liu et al.

Vehicle-to-everything (V2X) collaborative perception has emerged as a promising solution to address the limitations of single-vehicle perception systems. However, existing V2X datasets are limited in scope, diversity, and quality. To address these gaps, we present Mixed Signals, a comprehensive V2X dataset featuring 45.1k point clouds and 240.6k bounding boxes collected from three connected autonomous vehicles (CAVs) equipped with two different configurations of LiDAR sensors, plus a roadside unit with dual LiDARs. Our dataset provides point clouds and bounding box annotations across 10 classes, ensuring reliable data for perception training. We provide detailed statistical analysis on the quality of our dataset and extensively benchmark existing V2X methods on it. The Mixed Signals dataset is ready-to-use, with precise alignment and consistent annotations across time and viewpoints. Dataset website is available at https://mixedsignalsdataset.cs.cornell.edu/.

Andrew Caunes, Thierry Chateau, Vincent Fremont

3D semantic segmentation is essential for autonomous driving and road infrastructure analysis, but state-of-the-art 3D models suffer from severe domain shift when applied across datasets. We propose a multi-view projection framework for unsupervised domain adaptation (UDA). Our method aligns LiDAR scans into coherent 3D scenes and renders them from multiple virtual camera poses to generate large-scale synthetic 2D datasets (PC2D) in various modalities. An ensemble of 2D segmentation models is trained on these modalities, and during inference, hundreds of views per scene are processed, with logits back-projected to 3D using an occlusion-aware voting scheme to produce point-wise labels. These labels can be used directly or to fine-tune a 3D segmentation model in the target domain. We evaluate our approach in both Real-to-Real and Simulation-to-Real UDA, achieving state-of-the-art performance in the Real-to-Real setting. Furthermore, we show that our framework enables segmentation of rare classes, leveraging only 2D annotations for those classes while relying on 3D annotations for others in the source domain.

Haixin Sun, Minh-Quan Dao, Vincent Fremont

Event-based cameras can overpass frame-based cameras limitations for important tasks such as high-speed motion detection during self-driving cars navigation in low illumination conditions. The event cameras' high temporal resolution and high dynamic range, allow them to work in fast motion and extreme light scenarios. However, conventional computer vision methods, such as Deep Neural Networks, are not well adapted to work with event data as they are asynchronous and discrete. Moreover, the traditional 2D-encoding representation methods for event data, sacrifice the time resolution. In this paper, we first improve the 2D-encoding representation by expanding it into three dimensions to better preserve the temporal distribution of the events. We then propose 3D-FlowNet, a novel network architecture that can process the 3D input representation and output optical flow estimations according to the new encoding methods. A self-supervised training strategy is adopted to compensate the lack of labeled datasets for the event-based camera. Finally, the proposed network is trained and evaluated with the Multi-Vehicle Stereo Event Camera (MVSEC) dataset. The results show that our 3D-FlowNet outperforms state-of-the-art approaches with less training epoch (30 compared to 100 of Spike-FlowNet).

Minh-Quan Dao, Vincent Frémont

Multi-object tracking (MOT) is an integral part of any autonomous driving pipelines because itproduces trajectories which has been taken by other moving objects in the scene and helps predicttheir future motion. Thanks to the recent advances in 3D object detection enabled by deep learning,track-by-detection has become the dominant paradigm in 3D MOT. In this paradigm, a MOT systemis essentially made of an object detector and a data association algorithm which establishes track-to-detection correspondence. While 3D object detection has been actively researched, associationalgorithms for 3D MOT seem to settle at a bipartie matching formulated as a linear assignmentproblem (LAP) and solved by the Hungarian algorithm. In this paper, we adapt a two-stage dataassociation method which was successful in image-based tracking to the 3D setting, thus providingan alternative for data association for 3D MOT. Our method outperforms the baseline using one-stagebipartie matching for data association by achieving 0.587 AMOTA in NuScenes validation set.

Ruddy Théodose, Dieumet Denis, Thierry Chateau et al.

Current neural networks-based object detection approaches processing LiDAR point clouds are generally trained from one kind of LiDAR sensors. However, their performances decrease when they are tested with data coming from a different LiDAR sensor than the one used for training, i.e., with a different point cloud resolution. In this paper, R-AGNO-RPN, a region proposal network built on fusion of 3D point clouds and RGB images is proposed for 3D object detection regardless of point cloud resolution. As our approach is designed to be also applied on low point cloud resolutions, the proposed method focuses on object localization instead of estimating refined boxes on reduced data. The resilience to low-resolution point cloud is obtained through image features accurately mapped to Bird's Eye View and a specific data augmentation procedure that improves the contribution of the RGB images. To show the proposed network's ability to deal with different point clouds resolutions, experiments are conducted on both data coming from the KITTI 3D Object Detection and the nuScenes datasets. In addition, to assess its performances, our method is compared to PointPillars, a well-known 3D detection network. Experimental results show that even on point cloud data reduced by $80\%$ of its original points, our method is still able to deliver relevant proposals localization.

Salvador Dominguez, Gaëtan Garcia, Arnaud Hamon et al.

Obtaining a realistic and accurate model of the longitudinal dynamics is key for a good speed control of a self-driving car. It is also useful to simulate the longitudinal behavior of the vehicle with high fidelity. In this paper, a straightforward and generic method for obtaining the friction, braking and propulsion forces as a function of speed, throttle input and brake input is proposed. Experimental data is recorded during tests over the full speed range to estimate the forces, to which the corresponding curves are adjusted. A simple and direct balance of forces in the direction tangent to the ground is used to obtain an estimation of the real forces involved. Then a model composed of approximate spline curves that fit the results is proposed. Using splines to model the dynamic response has the advantage of being quick and accurate, avoiding the complexity of parameter identification and tuning of non-linear responses embedding the internal functionalities of the car, like ABS or regenerative brake. This methodology has been applied to LS2N's electric Renault Zoe but can be applied to any other electric car. As shown in the experimental section, a comparison between the estimated acceleration of the car using the model and the real one over a wide range of speeds along a trip of about $10km/h$ reveals only $0.35m/s^2$ of error standard deviation in a range of $\pm{2}m/s^2$ which is very encouraging.

Salvador Dominguez, Gaëtan Garcia, Vincent Frémont et al.

One of the hardest challenges to face in the development of a non GPS-based localization system for autonomous vehicles is the changes of the environment. LiDAR-based systems typically try to match the last measurements obtained with a previously recorded map of the area. If the existing map is not updated along time, there is a good chance that the measures will not match the environment well enough, causing the vehicle to lose track of its location. In this paper, we present and analyze experimental results regarding the robustness and precision of a map-matching based localization system over a certain period of time in the following three cases: (1) without any update of the initial map, (2) updating the map as the vehicle moves and (3) with map updates that take into account surrounding structures labeled as "fixed" which are treated differently. The environment of the tests is a busy parking area, which ensures drastic changes from one day to the next. The precision is obtained by comparing the positions computed using the map with the ones provided by a Real-Time Kinematic GPS system. The experimental results reveal a positioning error of about 6cm which remains stable even after 23 days when using fixed structures on the working area.

Ariane Spaenlehauer, Vincent Fremont, Y. Ahmet Sekercioglu et al.

In monocular vision systems, lack of knowledge about metric distances caused by the inherent scale ambiguity can be a strong limitation for some applications. We offer a method for fusing inertial measurements with monocular odometry or tracking to estimate metric distances in inertial-monocular systems and to increase the rate of pose estimates. As we performed the fusion in a loosely-coupled manner, each input block can be easily replaced with one's preference, which makes our method quite flexible. We experimented our method using the ORB-SLAM algorithm for the monocular tracking input and Euler forward integration to process the inertial measurements. We chose sets of data recorded on UAVs to design a suitable system for flying robots.

Xiaoqin Wang, Y. Ahmet Sekercioglu, Tom Drummond et al.

The Relative Pose based Redundancy Removal(RPRR) scheme is presented, which has been designed for mobile RGB-D sensor networks operating under bandwidth-constrained operational scenarios. Participating sensor nodes detect the redundant visual and depth information to prevent their transmission leading to a significant improvement in wireless channel usage efficiency and power savings. Experimental results show that wireless channel utilization is improved by 250% and battery consumption is halved when the RPRR scheme is used instead of sending the sensor images independently.

Caio César Teodoro Mendes, Vincent Frémont, Denis Fernando Wolf

Road detection is a fundamental task in autonomous navigation systems. In this paper, we consider the case of monocular road detection, where images are segmented into road and non-road regions. Our starting point is the well-known machine learning approach, in which a classifier is trained to distinguish road and non-road regions based on hand-labeled images. We proceed by introducing the use of "contextual blocks" as an efficient way of providing contextual information to the classifier. Overall, the proposed methodology, including its image feature selection and classifier, was conceived with computational cost in mind, leaving room for optimized implementations. Regarding experiments, we perform a sensible evaluation of each phase and feature subset that composes our system. The results show a great benefit from using contextual blocks and demonstrate their computational efficiency. Finally, we submit our results to the KITTI road detection benchmark achieving scores comparable with state of the art methods.

Paul George, Indira Thouvenin, Vincent Fremont et al.

Taking into account the drivers' state is a major challenge for designing new advanced driver assistance systems. In this paper we present a driver assistance system strongly coupled to the user. DAARIA 1 stands for Driver Assistance by Augmented Reality for Intelligent Automobile. It is an augmented reality interface powered by several sensors. The detection has two goals: one is the position of obstacles and the quantification of the danger represented by them. The other is the driver's behavior. A suitable visualization metaphor allows the driver to perceive at any time the location of the relevant hazards while keeping his eyes on the road. First results show that our method could be applied to a vehicle but also to aerospace, fluvial or maritime navigation.