Abstract--Ensuring reliable autonomous operation when visual input is degraded remains a key challenge in intelligent vehicles and robotics. We present DepthVision, a multimodal framework that enables Vision-Language Models (VLMs) to exploit LiDAR data without any architectural changes or retraining. DepthVision synthesizes dense, RGB-like images from sparse LiDAR point clouds using a conditional GAN with an integrated refiner, and feeds these into off-the-shelf VLMs through their standard visual interface. A Luminance-A ware Modality Adaptation (LAMA) module fuses synthesized and real camera images by dynamically weighting each modality based on ambient lighting, compensating for degradation such as darkness or motion blur . This design turns LiDAR into a drop-in visual surrogate when RGB becomes unreliable, effectively extending the operational envelope of existing VLMs. We evaluate DepthVision on real and simulated datasets across multiple VLMs and safety-critical tasks, including vehicle-in-the-loop experiments. The results show substantial improvements in low-light scene understanding over RGB-only baselines while preserving full compatibility with frozen VLM architectures. These findings demonstrate that LiDAR-guided RGB synthesis is a practical pathway for integrating range sensing into modern vision-language systems for autonomous driving. Intelligent vehicles and autonomous driving systems rely on accurate environment perception, prediction, and planning to ensure safe decision making and control. Classic autonomy stacks process raw sensor data through modular pipelines--covering perception, motion prediction, and trajectory planning--to reconstruct the scene and forecast dynamic agent behavior [1]-[3]. These modular systems have demonstrated high robustness across diverse driving scenarios, yet overall performance remains constrained by upstream sensing quality and cross-module coordination [4], [5].

Neural Information Processing Systems http://nips.cc/

LA VIB comprises a large collection of high-resolution videos sourced from the web through an automated pipeline with minimal requirements for human verification. Metrics are computed for each video's motion magnitudes, luminance conditions, frame

Ensuring safe operation of safety-critical complex systems interacting with their environment poses significant challenges, particularly when the system's world model relies on machine learning algorithms to process the perception input. A comprehensive safety argumentation requires knowledge of how faults or functional insufficiencies propagate through the system and interact with external factors, to manage their safety impact. While statistical analysis approaches can support the safety assessment, associative reasoning alone is neither sufficient for the safety argumentation nor for the identification and investigation of safety measures. A causal understanding of the system and its interaction with the environment is crucial for safeguarding safety-critical complex systems. It allows to transfer and generalize knowledge, such as insights gained from testing, and facilitates the identification of potential improvements. This work explores using causal Bayesian networks to model the system's causalities for safety analysis, and proposes measures to assess causal influences based on Pearl's framework of causal inference. We compare the approach of causal Bayesian networks to the well-established fault tree analysis, outlining advantages and limitations. In particular, we examine importance metrics typically employed in fault tree analysis as foundation to discuss suitable causal metrics. An evaluation is performed on the example of a perception system for automated driving. Overall, this work presents an approach for causal reasoning in safety analysis that enables the integration of data-driven and expert-based knowledge to account for uncertainties arising from complex systems operating in open environments.

Color constancy is the recovery of true surface color from observed color, and requires estimating the chromaticity of scene illumination to correct for the bias it induces. In this paper, we show that the per-pixel color statistics of natural scenes---without any spatial or semantic context---can by themselves be a powerful cue for color constancy. Specifically, we describe an illuminant estimation method that is built around a classifier for identifying the true chromaticity of a pixel given its luminance (absolute brightness across color channels). During inference, each pixel's observed color restricts its true chromaticity to those values that can be explained by one of a candidate set of illuminants, and applying the classifier over these values yields a distribution over the corresponding illuminants. A global estimate for the scene illuminant is computed through a simple aggregation of these distributions across all pixels.

Modern deep-learning based super-resolution techniques process images and videos independently of the underlying content and viewing conditions. However, the sensitivity of the human visual system to image details changes depending on the underlying content characteristics, such as spatial frequency, luminance, color, contrast, or motion. This observation hints that computational resources spent on up-sampling visual content may be wasted whenever a viewer cannot resolve the results. Motivated by this observation, we propose a perceptually inspired and architecture-agnostic approach for controlling the visual quality and efficiency of super-resolution techniques. The core is a perceptual model that dynamically guides super-resolution methods according to the human's sensitivity to image details. Our technique leverages the limitations of the human visual system to improve the efficiency of super-resolution techniques by focusing computational resources on perceptually important regions; judged on the basis of factors such as adapting luminance, contrast, spatial frequency, motion, and viewing conditions. We demonstrate the application of our proposed model in combination with network branching, and network complexity reduction to improve the computational efficiency of super-resolution methods without visible quality loss. Quantitative and qualitative evaluations, including user studies, demonstrate the effectiveness of our approach in reducing FLOPS by factors of 2$\mathbf{x}$ and greater, without sacrificing perceived quality.

Localizing target objects in images is an important task in computer vision. Often it is the first step towards solving a variety of applications in autonomous driving, maintenance, quality insurance, robotics, and augmented reality. Best in class solutions for this task rely on deep neural networks, which require a set of representative training data for best performance. Creating sets of sufficient quality, variety, and size is often difficult, error prone, and expensive. This is where the method of luminance keying [10,8] can help: it provides a simple yet effective solution to record high quality data for training object detection and segmentation. We extend previous work that presented luminance keying on the common YCB-V set of household objects [14] by recording the remaining objects of the YCB superset. The additional variety of objects - addition of transparency, multiple color variations, non-rigid objects - further demonstrates the usefulness of luminance keying and might be used to test the applicability of the approach on new 2D object detection and segmentation algorithms.

We present a practical distillation approach to fine-tune LLMs for invoking tools in real-time applications. We focus on visual editing tasks; specifically, we modify images and videos by interpreting user stylistic requests, specified in natural language ("golden hour"), using an LLM to select the appropriate tools and their parameters to achieve the desired visual effect. We found that proprietary LLMs such as GPT-3.5-Turbo show potential in this task, but their high cost and latency make them unsuitable for real-time applications. In our approach, we fine-tune a (smaller) student LLM with guidance from a (larger) teacher LLM and behavioral signals. We introduce offline metrics to evaluate student LLMs. Both online and offline experiments show that our student models manage to match the performance of our teacher model (GPT-3.5-Turbo), significantly reducing costs and latency. Lastly, we show that fine-tuning was improved by 25% in low-data regimes using augmentation.

Recent advancements in vision models have greatly improved their ability to handle complex chart understanding tasks, like chart captioning and question answering. However, it remains challenging to assess how these models process charts. Existing benchmarks only roughly evaluate model performance without evaluating the underlying mechanisms, such as how models extract image embeddings. This limits our understanding of the model's ability to perceive fundamental graphical components. To address this, we introduce a novel evaluation framework to assess the graphical perception of image embedding models. For chart comprehension, we examine two main aspects of channel effectiveness: accuracy and discriminability of various visual channels. Channel accuracy is assessed through the linearity of embeddings, measuring how well the perceived magnitude aligns with the size of the stimulus. Discriminability is evaluated based on the distances between embeddings, indicating their distinctness. Our experiments with the CLIP model show that it perceives channel accuracy differently from humans and shows unique discriminability in channels like length, tilt, and curvature. We aim to develop this work into a broader benchmark for reliable visual encoders, enhancing models for precise chart comprehension and human-like perception in future applications.