Explore the vast range of titles available.

This survey reviews explainability methods for vision-based self-driving systems trained with behavior cloning. The concept of explainability has several facets and the need for explainability is strong in driving, a safety-critical application. Gathering contributions from several research fields, namely computer vision, deep learning, autonomous driving, explainable AI (X-AI), this survey tackles several points. First, it discusses definitions, context, and motivation for gaining more interpretability and explainability from self-driving systems, as well as the challenges that are specific to this application. Second, methods providing explanations to a black-box self-driving system in a post-hoc fashion are comprehensively organized and detailed. Third, approaches from the literature that aim at building more interpretable self-driving systems by design are presented and discussed in detail. Finally, remaining open-challenges and potential future research directions are identified and examined.

Object detection methods usually represent objects through rectangular bounding boxes from which they extract features, regardless of their actual shapes. In this paper, we apply deformations to regions in order to learn representations better fitted to objects. We introduce DP-FCN, a deep model implementing this idea by learning to align parts to discriminative elements of objects in a latent way, i.e. without part annotation. This approach has two main assets: it builds invariance to local transformations, thus improving recognition, and brings geometric information to describe objects more finely, leading to a more accurate localization. We further develop both features in a new model named DP-FCN2.0 by explicitly learning interactions between parts. Alignment is done with an in-network joint optimization of all parts based on a CRF with custom potentials, and deformations are influencing localization through a bilinear product. We validate our models on PASCAL VOC and MS COCO datasets and show significant gains. DP-FCN2.0 achieves state-of-the-art results of 83.3 and 81.2% on VOC 2007 and 2012 with VOC data only.

This paper is concerned with the problem of learning a distance metric by considering meaningful and discriminative distance constraints in some contexts where rich information between data is provided. Classic metric learning approaches focus on constraints that involve pairs or triplets of images. We propose a general Mahalanobis-like distance metric learning framework that exploits distance constraints over up to four different images. We show how the integration of such constraints can lead to unsupervised or semi-supervised learning tasks in some applications. We also show the benefit on recognition performance of this type of constraints, in rich contexts such as relative attributes, class taxonomies and temporal webpage analysis.

In this article, we propose a new video object retrieval system. Our approach is based on a Spatio-Temporal data representation, a dedicated kernel design and a statistical learning toolbox for video object recognition and retrieval. Using state-of-the-art video object detection algorithms (for faces or cars, for example) we segment video object tracks from real movies video shots. We then extract, from these tracks, sets of spatio-temporally coherent features that we call Spatio-Temporal Tubes. To compare these complex tube objects, we design a Spatio-Temporal Tube Kernel (STTK) function. Based on this kernel similarity we present both supervised and active learning strategies embedded in Support Vector Machine framework. Additionally, we propose a multi-class classification framework dealing with unbalanced data. Our approach is successfully evaluated on two real movies databases, the french movie “L’esquive” and episodes from “Buffy, the Vampire Slayer” TV series. Our method is also tested on a car database (from real movies) and shows promising results for car identification task.

Large Language Models (LLMs) have demonstrated remarkable success in both textual and multimodal domains. However, this success often comes with substantial computational costs, particularly when handling lengthy sequences of multimodal inputs. This has sparked many efforts focusing on enhancing efficiency during training and inference. In this study, we investigate the computation redundancy in Multimodal Large Language Models (MLLMs) during inference. We propose different methods to skip computations, such as skipping entire blocks, FFN or self-attention (SA) layers. Additionally, we explore parallelizing certain layers, such as FFN and SA layers. Our findings validate that (1) significant amount of computations can be avoided at inference time, especially for tasks such as Visual Question Answering (VQA). (2) Skipping computations during training can recover 97% of the original performance, even when skipping half of the blocks or removing 70% of the weights. Alternatively, (3) properly training with smaller LLMs can yield comparable performance to LLMs 2 or 3 times larger. To conclude, we extend our investigation to recent MLLMs, such as LLaVA-1.5, showing similar observations. Our work show that there is redundant computations inside MLLMs and thus the potential for significantly improving inference costs without sacrificing performance. The code is available here: https://github.com/mshukor/ima-lmms.

We consider the problem of text-to-video generation tasks with precise control for various applications such as camera movement control and video-to-video editing. Most methods tacking this problem rely on providing user-defined controls, such as binary masks or camera movement embeddings. In our approach we propose OnlyFlow, an approach leveraging the optical flow firstly extracted from an input video to condition the motion of generated videos. Using a text prompt and an input video, OnlyFlow allows the user to generate videos that respect the motion of the input video as well as the text prompt. This is implemented through an optical flow estimation model applied on the input video, which is then fed to a trainable optical flow encoder. The output feature maps are then injected into the text-to-video backbone model. We perform quantitative, qualitative and user preference studies to show that OnlyFlow positively compares to state-of-the-art methods on a wide range of tasks, even though OnlyFlow was not specifically trained for such tasks. OnlyFlow thus constitutes a versatile, lightweight yet efficient method for controlling motion in text-to-video generation. Models and code will be made available on GitHub and HuggingFace.

Transformers have been matching deep convolutional networks for vision architectures in recent works. Most work is focused on getting the best results on large-scale benchmarks, and scaling laws seem to be the most successful strategy: bigger models, more data, and longer training result in higher performance. However, the reduction of network complexity and inference time remains under-explored. The Perceiver model offers a solution to this problem: by first performing a Cross-attention with a fixed number Q of latent query tokens, the complexity of the L-layers Transformer network that follows is bounded by O(LQ^2). In this work, we explore how to make Perceivers even more efficient, by reducing the number of queries Q during inference while limiting the accuracy drop.

Tokenizers are a key component of state-of-the-art generative image models, extracting the most important features from the signal while reducing data dimension and redundancy. Most current tokenizers are based on KL-regularized variational autoencoders (KL-VAE), trained with reconstruction, perceptual and adversarial losses. Diffusion decoders have been proposed as a more principled alternative to model the distribution over images conditioned on the latent. However, matching the performance of KL-VAE still requires adversarial losses, as well as a higher decoding time due to iterative sampling. To address these limitations, we introduce a new pixel diffusion decoder architecture for improved scaling and training stability, benefiting from transformer components and GAN-free training. We use distillation to replicate the performance of the diffusion decoder in an efficient single-step decoder. This makes SSDD the first diffusion decoder optimized for single-step reconstruction trained without adversarial losses, reaching higher reconstruction quality and faster sampling than KL-VAE. In particular, SSDD improves reconstruction FID from \\(0.87\\) to \\(0.50\\) with \\(1.4\\) higher throughput and preserve generation quality of DiTs with \\(3.8\\) faster sampling. As such, SSDD can be used as a drop-in replacement for KL-VAE, and for building higher-quality and faster generative models.

Existing vehicle trajectory prediction models struggle with generalizability, prediction uncertainties, and handling complex interactions. It is often due to limitations like complex architectures customized for a specific dataset and inefficient multimodal handling. We propose Perceiver with Register queries (PerReg+), a novel trajectory prediction framework that introduces: (1) Dual-Level Representation Learning via Self-Distillation (SD) and Masked Reconstruction (MR), capturing global context and fine-grained details. Additionally, our approach of reconstructing segmentlevel trajectories and lane segments from masked inputs with query drop, enables effective use of contextual information and improves generalization; (2) Enhanced Multimodality using register-based queries and pretraining, eliminating the need for clustering and suppression; and (3) Adaptive Prompt Tuning during fine-tuning, freezing the main architecture and optimizing a small number of prompts for efficient adaptation. PerReg+ sets a new state-of-the-art performance on nuScenes [1], Argoverse 2 [2], and Waymo Open Motion Dataset (WOMD) [3]. Remarkable, our pretrained model reduces the error by 6.8% on smaller datasets, and multi-dataset training enhances generalization. In cross-domain tests, PerReg+ reduces B-FDE by 11.8% compared to its non-pretrained variant.

Recent video diffusion models generate photorealistic, temporally coherent videos, yet they fall short as reliable world models for autonomous driving, where structured motion and physically consistent interactions are essential. Adapting these generalist video models to driving domains has shown promise but typically requires massive domain-specific data and costly fine-tuning. We propose an efficient adaptation framework that converts generalist video diffusion models into controllable driving world models with minimal supervision. The key idea is to decouple motion learning from appearance synthesis. First, the model is adapted to predict structured motion in a simplified form: videos of skeletonized agents and scene elements, focusing learning on physical and social plausibility. Then, the same backbone is reused to synthesize realistic RGB videos conditioned on these motion sequences, effectively \"dressing\" the motion with texture and lighting. This two-stage process mirrors a reasoning-rendering paradigm: first infer dynamics, then render appearance. Our experiments show this decoupled approach is exceptionally efficient: adapting SVD, we match prior SOTA models with less than 6% of their compute. Scaling to LTX, our MAD-LTX model outperforms all open-source competitors, and supports a comprehensive suite of text, ego, and object controls. Project page: https://vita-epfl.github.io/MAD-World-Model/