Searcharxiv⌕ Search

arXiv subjects

Abdalmalek Aburaddaha

Publications and source records attributed to Abdalmalek Aburaddaha.

4 recordsLinked to original sources

From Cloud to Edge: A Fully Local Pipeline for Pothole Risk Assessment and Driver Warning Generation

Vision-language models can turn road images into clear, geometry-aware pothole warnings, but current systems depend on commercial cloud services. Our earlier five-stage pipeline needed two cloud calls per scene, taking about 20 s and USD 0.105, and could not run offline. We rebuild every stage to run on a single local GPU and evaluate it with the same data, prompts, and protocol. Segmentation keeps the original Segment Anything weights but runs them in half precision with graph compilation, which makes it 2.7x faster without changing any downstream risk class; five lightweight variants could not match its masks on small road defects. Severity is graded by six ConvNeXt students distilled from 3092 cloud labels. They reach Cohen's kappa = 0.268 against human consensus at 28 ms per pothole, close to the cloud model's 0.321, within the camera setup they were trained for. A local 8-billion-parameter vision-language model writes the warnings under a grammar that admits only in-lane maneuvers. Three blind, independent raters scored these warnings safer than the cloud warnings (4.18 vs. 3.65 of 5; p = 0.001; 50 scenes) and equally clear, whereas an automated LLM judge ranked them the other way. The complete pipeline runs at 1.27 s per scene, 16x faster than the cloud system, offline and at no per-scene cost.

cs.CV↗

Match the Amplitude, Not the Layout: Phase-Aware Distillation from RGB Foundation Models to Infrared Detection

Cross-modal distillation measures a single similarity per frequency band, observes that the high band is near zero, and concludes that it should be aligned loosely. That inference conflates the kind of disagreement with its degree. We decompose the agreement between a frozen RGB foundation teacher and an infrared (IR) student into two components that a scalar similarity confounds: phase, which encodes the spatial position of structure, and amplitude, which encodes its magnitude. Measured on several hundred registered pairs prior to any training, the two components diverge: for periods finer than 60 px amplitude agreement remains at 0.52, while phase coherence falls to 0.06, indistinguishable from chance. What ceases to transfer at fine scales is spatial location, not magnitude. The high band, therefore, requires a different quantity rather than a smaller weight: we supervise the low band as a complex-valued feature and the high band in amplitude alone. The cut-off is not tuned on accuracy but derived as the radius at which teacher content ceases to be linearly recoverable from the student, yielding rc=0.47. A single distilled IR backbone, reused without re-distillation, attains 64.1 mAP50 on KAIST Person (+2.4 over the DINOv2 baseline) and initializes FLIR detection (+2.1 mAP50) and MFNet segmentation (+1.85 mIoU); the same objective also improves a ResNet-50 student (+1.0 mAP50). All reported results use a DINOv2 ViT-Large teacher. Code and the distilled backbone are available at github.com/tkeval/freq_decoupled_kd.

cs.CV↗

Pothole Risk Assessment with Geometrically Grounded Vision Language Models

Potholes impose billions of dollars in annual vehicle damage, yet detectors report only where potholes are, not how severe they are or what to do. Judging severity from a single RGB image is ill-posed, and ungrounded vision-language model (VLM) assessments tend to be over-conservative. We present a five-stage modular framework for the offline assessment of already-detected potholes that combines geometric measurements with VLM reasoning: perspective transformation, zero-shot SAM segmentation, geometric feature extraction, Gemini 2.5 Pro severity classification with composite risk scoring, and natural-language warning generation. We evaluate it on 3,092 potholes in 1,119 dashcam scenes against independent references. Severity agreement with a prediction-free three-annotator panel is weak (Cohen's $κ= 0.139$, 95% CI [0.00, 0.28]; $n = 101$) and far below the annotators' agreement with one another ($κ= 0.605$); against physically measured depth on an independent RGB-D dataset, it is also weak ($κ= 0.098$), although the predicted classes preserve the order of measured depths. The severity classes are therefore perceived visual categories, not depth measurements. Geometric grounding does change the warnings: relative to an image-only baseline, it moves 26.7% of scenes out of the high-urgency tier and raises critical-tier assignments by 109% (Stuart-Maxwell $p < 10^{-15}$), and in a blind study, three raters preferred the grounded warnings 3.5:1. A six-arm prompt ablation attributes the hybrid prompt's rule adherence to the per-pothole measurements rather than to the rubric or worked examples. At USD 0.105 per scene, it suits large-scale offline infrastructure assessment; its maneuver suggestions are advisory research outputs, not validated vehicle guidance.

cs.CV↗

SceneMiner: Identity-Preserving Multi-Task Fine-Tuning for Unified BEV Scene Mining

Mining hard, safety-critical scenes from driving logs is bottlenecked by the absence of difficulty labels, and no single proxy, collision risk, trajectory ambiguity, or semantic rarity suffices to find such scenes on its own. We present SceneMiner, a unified, camera-only bird's-eye-view pipeline that emits complementary mining signals from a frozen vision-language backbone in a single forward pass, with no LiDAR or radar: a retrieval embedding for text-prompted scenario search, a multi-label scene-tag distribution, and a continuous physics-based risk score (a motion forecast is a byproduct, not a contribution). Building such a multi-head model exposes our central finding, a failure mode we term cross-task interference: adding or upgrading one head shifts a shared activation stream and degrades weight-frozen sibling heads, so freezing parameters alone is insufficient. Our contribution, identity-preserving multi-task fine-tuning, removes this interference by zero-initializing every new sub-module and freezing every parameter that feeds the shared stream. The mining heads are thereby preserved bit-identically while training only ~102k parameters. The tagging head reaches mAP 0.4614 (micro-F1 0.5557) on 20 scene tags by pooling each scene into 32 visual tokens, and the embedding head supports text-prompted retrieval, validated qualitatively. Code is available at: https://anonymous.4open.science/r/sceneminer_anonymous-64E5

cs.CV↗