Annealed Winner-Takes-All for Motion Forecasting

In autonomous driving, motion prediction aims at forecasting the future trajectories of nearby agents, helping the ego vehicle to anticipate behaviors and drive safely. A key challenge is generating a diverse set of future predictions, commonly addressed using data-driven models with Multiple Choice Learning (MCL) architectures and Winner-Takes-All (WTA) training objectives. However, these methods face initialization sensitivity and training instabilities. Additionally, to compensate for limited performance, some approaches rely on training with a large set of hypotheses, requiring a post-selection step during inference to significantly reduce the number of predictions. To tackle these issues, we take inspiration from annealed MCL, a recently introduced technique that improves the convergence properties of MCL methods through an annealed Winner-Takes-All loss (aWTA). In this paper, we demonstrate how the aWTA loss can be integrated with state-of-the-art motion forecasting models to enhance their performance using only a minimal set of hypotheses, eliminating the need for the cumbersome post-selection step. Our approach can be easily incorporated into any trajectory prediction model normally trained using WTA and yields significant improvements. To facilitate the application of our approach to future motion forecasting models, the code will be made publicly available upon acceptance: https://github.com/valeoai/MF_aWTA.

Lost and Found: Overcoming Detector Failures in Online Multi-Object Tracking

Multi-object tracking (MOT) endeavors to precisely estimate the positions and identities of multiple objects over time. The prevailing approach, tracking-by-detection (TbD), first detects objects and then links detections, resulting in a simple yet effective method. However, contemporary detectors may occasionally miss some objects in certain frames, causing trackers to cease tracking prematurely. To tackle this issue, we propose BUSCA, meaning `to search', a versatile framework compatible with any online TbD system, enhancing its ability to persistently track those objects missed by the detector, primarily due to occlusions. Remarkably, this is accomplished without modifying past tracking results or accessing future frames, i.e., in a fully online manner. BUSCA generates proposals based on neighboring tracks, motion, and learned tokens. Utilizing a decision Transformer that integrates multimodal visual and spatiotemporal information, it addresses the object-proposal association as a multi-choice question-answering task. BUSCA is trained independently of the underlying tracker, solely on synthetic data, without requiring fine-tuning. Through BUSCA, we showcase consistent performance enhancements across five different trackers and establish a new state-of-the-art baseline across three different benchmarks. Code available at: https://github.com/lorenzovaquero/BUSCA.

Valeo4Cast: A Modular Approach to End-to-End Forecasting

Motion forecasting is crucial in autonomous driving systems to anticipate the future trajectories of surrounding agents such as pedestrians, vehicles, and traffic signals. In end-to-end forecasting, the model must jointly detect from sensor data (cameras or LiDARs) the position and past trajectories of the different elements of the scene and predict their future location. We depart from the current trend of tackling this task via end-to-end training from perception to forecasting and we use a modular approach instead. Following a recent study, we individually build and train detection, tracking, and forecasting modules. We then only use consecutive finetuning steps to integrate the modules better and alleviate compounding errors. Our study reveals that this simple yet effective approach significantly improves performance on the end-to-end forecasting benchmark. Consequently, our solution ranks first in the Argoverse 2 end-to-end Forecasting Challenge held at CVPR 2024 Workshop on Autonomous Driving (WAD), with 63.82 mAPf. We surpass forecasting results by +17.1 points over last year's winner and by +13.3 points over this year's runner-up. This remarkable performance in forecasting can be explained by our modular paradigm, which integrates finetuning strategies and significantly outperforms the end-to-end-trained counterparts.

Learning Kernel-Modulated Neural Representation for Efficient Light Field Compression

Light field is a type of image data that captures the 3D scene information by recording light rays emitted from a scene at various orientations. It offers a more immersive perception than classic 2D images but at the cost of huge data volume. In this paper, we draw inspiration from the visual characteristics of Sub-Aperture Images (SAIs) of light field and design a compact neural network representation for the light field compression task. The network backbone takes randomly initialized noise as input and is supervised on the SAIs of the target light field. It is composed of two types of complementary kernels: descriptive kernels (descriptors) that store scene description information learned during training, and modulatory kernels (modulators) that control the rendering of different SAIs from the queried perspectives. To further enhance compactness of the network meanwhile retain high quality of the decoded light field, we accordingly introduce modulator allocation and kernel tensor decomposition mechanisms, followed by non-uniform quantization and lossless entropy coding techniques, to finally form an efficient compression pipeline. Extensive experiments demonstrate that our method outperforms other state-of-the-art (SOTA) methods by a significant margin in the light field compression task. Moreover, after aligning descriptors, the modulators learned from one light field can be transferred to new light fields for rendering dense views, indicating a potential solution for view synthesis task.

Challenges of Using Real-World Sensory Inputs for Motion Forecasting in Autonomous Driving

Motion forecasting plays a critical role in enabling robots to anticipate future trajectories of surrounding agents and plan accordingly. However, existing forecasting methods often rely on curated datasets that are not faithful to what real-world perception pipelines can provide. In reality, upstream modules that are responsible for detecting and tracking agents, and those that gather road information to build the map, can introduce various errors, including misdetections, tracking errors, and difficulties in being accurate for distant agents and road elements. This paper aims to uncover the challenges of bringing motion forecasting models to this more realistic setting where inputs are provided by perception modules. In particular, we quantify the impacts of the domain gap through extensive evaluation. Furthermore, we design synthetic perturbations to better characterize their consequences, thus providing insights into areas that require improvement in upstream perception modules and guidance toward the development of more robust forecasting methods.

Learning-based Spatial and Angular Information Separation for Light Field Compression

Light fields are a type of image data that capture both spatial and angular scene information by recording light rays emitted by a scene from different orientations. In this context, spatial information is defined as features that remain static regardless of perspectives, while angular information refers to features that vary between viewpoints. We propose a novel neural network that, by design, can separate angular and spatial information of a light field. The network represents spatial information using spatial kernels shared among all Sub-Aperture Images (SAIs), and angular information using sets of angular kernels for each SAI. To further improve the representation capability of the network without increasing parameter number, we also introduce angular kernel allocation and kernel tensor decomposition mechanisms. Extensive experiments demonstrate the benefits of information separation: when applied to the compression task, our network outperforms other state-of-the-art methods by a large margin. And angular information can be easily transferred to other scenes for rendering dense views, showing the successful separation and the potential use case for the view synthesis task. We plan to release the code upon acceptance of the paper to encourage further research on this topic.

DNN Training Acceleration via Exploring GPGPU Friendly Sparsity

The training phases of Deep neural network~(DNN) consumes enormous processing time and energy. Compression techniques utilizing the sparsity of DNNs can effectively accelerate the inference phase of DNNs. However, it is hardly used in the training phase because the training phase involves dense matrix-multiplication using General-Purpose Computation on Graphics Processors (GPGPU), which endorse the regular and structural data layout. In this paper, we first propose the Approximate Random Dropout that replaces the conventional random dropout of neurons and synapses with a regular and online generated row-based or tile-based dropout patterns to eliminate the unnecessary computation and data access for the multilayer perceptron~(MLP) and long short-term memory~(LSTM). We then develop a SGD-based Search Algorithm that produces the distribution of row-based or tile-based dropout patterns to compensate for the potential accuracy loss. Moreover, aiming at the convolution neural network~(CNN) training acceleration, we first explore the importance and sensitivity of input feature maps; and then propose the sensitivity-aware dropout method to dynamically drop the input feature maps based on their sensitivity so as to achieve greater forward and backward training acceleration while reserving better NN accuracy. To facilitate DNN programming, we build a DNN training computation framework that unifies the proposed techniques in the software stack. As a result, the GPGPU only needs to support the basic operator -- matrix multiplication and can achieve significant performance improvement regardless of DNN model.

CP-ViT: Cascade Vision Transformer Pruning via Progressive Sparsity Prediction

Vision transformer (ViT) has achieved competitive accuracy on a variety of computer vision applications, but its computational cost impedes the deployment on resource-limited mobile devices. We explore the sparsity in ViT and observe that informative patches and heads are sufficient for accurate image recognition. In this paper, we propose a cascade pruning framework named CP-ViT by predicting sparsity in ViT models progressively and dynamically to reduce computational redundancy while minimizing the accuracy loss. Specifically, we define the cumulative score to reserve the informative patches and heads across the ViT model for better accuracy. We also propose the dynamic pruning ratio adjustment technique based on layer-aware attention range. CP-ViT has great general applicability for practical deployment, which can be applied to a wide range of ViT models and can achieve superior accuracy with or without fine-tuning. Extensive experiments on ImageNet, CIFAR-10, and CIFAR-100 with various pre-trained models have demonstrated the effectiveness and efficiency of CP-ViT. By progressively pruning 50\% patches, our CP-ViT method reduces over 40\% FLOPs while maintaining accuracy loss within 1\%.

TransCenter: Transformers with Dense Queries for Multiple-Object Tracking

Transformer networks have proven extremely powerful for a wide variety of tasks since they were introduced. Computer vision is not an exception, as the use of transformers has become very popular in the vision community in recent years. Despite this wave, multiple-object tracking (MOT) exhibits for now some sort of incompatibility with transformers. We argue that the standard representation -- bounding boxes -- is not adapted to learning transformers for MOT. Inspired by recent research, we propose TransCenter, the first transformer-based architecture for tracking the centers of multiple targets. Methodologically, we propose the use of dense queries in a double-decoder network, to be able to robustly infer the heatmap of targets' centers and associate them through time. TransCenter outperforms the current state-of-the-art in multiple-object tracking, both in MOT17 and MOT20. Our ablation study demonstrates the advantage in the proposed architecture compared to more naive alternatives. The code will be made publicly available.

DeepMOT: A Differentiable Framework for Training Multiple Object Trackers

Multiple Object Tracking accuracy and precision (MOTA and MOTP) are two standard and widely-used metrics to assess the quality of multiple object trackers. They are specifically designed to encode the challenges and difficulties of tracking multiple objects. To directly optimize a tracker based on MOTA and MOTP is difficult, since both the metrics are strongly rely on the Hungarian algorithm, which are non-differentiable. We propose a differentiable proxy for the MOTA and MOTP, thus allowing to train a deep multiple-object tracker by directly optimizing (a proxy of) the standard MOT metrics. The proposed approximation is based on a bidirectional recurrent network that inputs the object-to-hypothesis distance matrix and outputs the optimal hypothesis-to-object association, thus emulating the Hungarian algorithm. Followed by a differentiable module, the estimated association is used to compute the MOTA and MOTP. The experimental study demonstrates the benefits of this differentiable framework on two recent deep trackers over the MOT17 dataset. Moreover, the code is publicly available from https://gitlab.inria.fr/yixu/deepmot.