Level 5 autonomy for self-driving cars requires a robust visual perception system that can parse input images under any visual condition. However, existing semantic segmentation datasets are either dominated by images captured under normal conditions or are small in scale. To address this, we introduce ACDC, the Adverse Conditions Dataset with Correspondences for training and testing semantic segmentation methods on adverse visual conditions. ACDC consists of a large set of 4006 images which are equally distributed between four common adverse conditions: fog, nighttime, rain, and snow. Each adverse-condition image comes with a high-quality fine pixel-level semantic annotation, a corresponding image of the same scene taken under normal conditions, and a binary mask that distinguishes between intra-image regions of clear and uncertain semantic content. Thus, ACDC supports both standard semantic segmentation and the newly introduced uncertainty-aware semantic segmentation. A detailed empirical study demonstrates the challenges that the adverse domains of ACDC pose to state-of-the-art supervised and unsupervised approaches and indicates the value of our dataset in steering future progress in the field. Our dataset and benchmark are publicly available.
We introduce AnyUp, a method for feature upsampling that can be applied to any vision feature at any resolution, without encoder-specific training. Existing learning-based upsamplers for features like DINO or CLIP need to be re-trained for every feature extractor and thus do not generalize to different feature types at inference time. In this work, we propose an inference-time feature-agnostic upsampling architecture to alleviate this limitation and improve upsampling quality. In our experiments, AnyUp sets a new state of the art for upsampled features, generalizes to different feature types, and preserves feature semantics while being efficient and easy to apply to a wide range of downstream tasks.
The development of multimodal large language models (MLLMs) enables the evaluation of image quality through natural language descriptions. This advancement allows for more detailed assessments. However, these MLLM-based IQA methods primarily rely on general contextual descriptions, sometimes limiting fine-grained quality assessment. To address this limitation, we introduce a new image quality assessment (IQA) task paradigm, **grounding-IQA**. This paradigm integrates multimodal referring and grounding with IQA to realize more fine-grained quality perception, thereby extending existing IQA. Specifically, grounding-IQA comprises two subtasks: grounding-IQA-description (GIQA-DES) and visual question answering (GIQA-VQA). GIQA-DES involves detailed descriptions with precise locations (e.g., bounding boxes), while GIQA-VQA focuses on quality QA for local regions. To realize grounding-IQA, we construct a corresponding dataset, GIQA-160K, through our proposed automated annotation pipeline. Furthermore, we develop a well-designed benchmark, GIQA-Bench. The benchmark evaluates the grounding-IQA performance from three perspectives: description quality, VQA accuracy, and grounding precision. Experiments demonstrate that our proposed method facilitates the more fine-grained IQA application. Code: https://github.com/zhengchen1999/Grounding-IQA.
In this work, we study the limits of compressed data structures, i.e., structures that support various queries on an input text $T\inΣ^n$ using space proportional to the size of $T$ in compressed form. Nearly all fundamental queries can currently be efficiently supported in $O(δ(T)\log^{O(1)}n)$ space, where $δ(T)$ is the substring complexity, a strong compressibility measure that lower-bounds the optimal space to represent the text [Kociumaka, Navarro, Prezza, IEEE Trans. Inf. Theory 2023]. However, optimal query time has been characterized only for random access. We address this gap by developing tight lower bounds for nearly all other fundamental queries: (1) We prove that suffix array (SA), inverse suffix array (SA$^{-1}$), longest common prefix (LCP) array, and longest common extension (LCE) queries all require $Ω(\log n/\log\log n)$ time within $O(δ(T)\log^{O(1)}n)$ space, matching known upper bounds. (2) We further show that other common queries, currently supported in $O(\log\log n)$ time and $O(δ(T)\log^{O(1)}n)$ space, including the Burrows-Wheeler Transform (BWT), permuted longest common prefix (PLCP) array, Last-to-First (LF), inverse LF, lexicographic predecessor ($Φ$), and inverse $Φ$ queries, all require $Ω(\log\log n)$ time, yielding another set of tight bounds. Our lower bounds hold even for texts over a binary alphabet. This work establishes a clean dichotomy: the optimal time complexity to support central string queries in compressed space is either $Θ(\log n/\log\log n)$ or $Θ(\log\log n)$. This completes the theoretical foundation of compressed indexing, closing a crucial gap between upper and lower bounds and providing a clear target for future data structures: seeking either the optimal time in the smallest space or the fastest time in the optimal space, both of which are now known for central string queries.
Radiance fields have emerged as a predominant representation for modeling 3D scene appearance. Neural formulations such as Neural Radiance Fields provide high expressivity but require costly ray marching for rendering, whereas primitive-based methods such as 3D Gaussian Splatting offer real-time efficiency through splatting, yet at the expense of representational power. Inspired by advances in both these directions, we introduce splattable neural primitives, a new volumetric representation that reconciles the expressivity of neural models with the efficiency of primitive-based splatting. Each primitive encodes a bounded neural density field parameterized by a shallow neural network. Our formulation admits an exact analytical solution for line integrals, enabling efficient computation of perspectively accurate splatting kernels. As a result, our representation supports integration along view rays without the need for costly ray marching. The primitives flexibly adapt to scene geometry and, being larger than prior analytic primitives, reduce the number required per scene. On novel-view synthesis benchmarks, our approach matches the quality and speed of 3D Gaussian Splatting while using 10x fewer primitives and 6x fewer parameters. These advantages arise directly from the representation itself, without reliance on complex control or adaptation frameworks.
