arXiv:2512.02258v1 Announce Type: new 
Abstract: Biologically plausible and energy-efficient frameworks such as Spiking Neural Networks (SNNs) have not been sufficiently explored in low-level vision tasks. Taking image deraining as an example, this study addresses the representation of the inherent high-pass characteristics of spiking neurons, specifically in image deraining and innovatively proposes the Visual LIF (VLIF) neuron, overcoming the obstacle of lacking spatial contextual understanding present in traditional spiking neurons. To tackle the limitation of frequency-domain saturation inherent in conventional spiking neurons, we leverage the proposed VLIF to introduce the Spiking Decomposition and Enhancement Module and the lightweight Spiking Multi-scale Unit for hierarchical multi-scale representation learning. Extensive experiments across five benchmark deraining datasets demonstrate that our approach significantly outperforms state-of-the-art SNN-based deraining methods, achieving this superior performance with only 13\% of their energy consumption. These findings establish a solid foundation for deploying SNNs in high-performance, energy-efficient low-level vision tasks.

دراسة حديثة استكشفت إمكانيات الشبكات العصبية المتفجرة (SNNs) في إزالة المطر من الصور، مقدمةً نيوترون Visual LIF (VLIF) لتحسين الفهم السياقي المكاني والتغلب على قيود النيوترونات المتفجرة التقليدية. تُظهر الأبحاث أن الأساليب المقترحة تتفوق بشكل كبير على تقنيات إزالة المطر المعتمدة على SNNs الموجودة عبر خمسة مجموعات بيانات مرجعية.

Un estudio reciente ha explorado los potenciales de las Redes Neuronales Espigadas (SNN) en la eliminación de lluvia en imágenes, introduciendo la neurona Visual LIF (VLIF) para mejorar la comprensión contextual espacial y superar las limitaciones de las neuronas espigadas tradicionales. La investigación demuestra que los métodos propuestos superan significativamente las técnicas de eliminación de lluvia basadas en SNN existentes en cinco conjuntos de datos de referencia.

Une étude récente a exploré les potentiels des réseaux de neurones à spikes (SNN) dans la détection de pluie sur les images, introduisant le neurone Visual LIF (VLIF) pour améliorer la compréhension contextuelle spatiale et surmonter les limitations des neurones à spikes traditionnels. La recherche démontre que les méthodes proposées surpassent significativement les techniques de détection de pluie basées sur SNN existantes à travers cinq ensembles de données de référence.

A recent study has explored the potentials of Spiking Neural Networks (SNNs) in image deraining, introducing the Visual LIF (VLIF) neuron to enhance spatial contextual understanding and overcome limitations of traditional spiking neurons. The research demonstrates that the proposed methods significantly outperform existing SNN-based deraining techniques across five benchmark datasets.

Exploring the Potentials of Spiking Neural Networks for Image Deraining

arXiv:2512.02447v1 Announce Type: new 
Abstract: Spiking Neural Networks (SNNs), with their brain-inspired spatiotemporal dynamics and spike-driven computation, have emerged as promising energy-efficient alternatives to Artificial Neural Networks (ANNs). However, existing SNNs typically replicate inputs directly or aggregate them into frames at fixed intervals. Such strategies lead to neurons receiving nearly identical stimuli across time steps, severely limiting the model's expressive power, particularly in complex tasks like object detection. In this work, we propose the Temporal Dynamics Enhancer (TDE) to strengthen SNNs' capacity for temporal information modeling. TDE consists of two modules: a Spiking Encoder (SE) that generates diverse input stimuli across time steps, and an Attention Gating Module (AGM) that guides the SE generation based on inter-temporal dependencies. Moreover, to eliminate the high-energy multiplication operations introduced by the AGM, we propose a Spike-Driven Attention (SDA) to reduce attention-related energy consumption. Extensive experiments demonstrate that TDE can be seamlessly integrated into existing SNN-based detectors and consistently outperforms state-of-the-art methods, achieving mAP50-95 scores of 57.7% on the static PASCAL VOC dataset and 47.6% on the neuromorphic EvDET200K dataset. In terms of energy consumption, the SDA consumes only 0.240 times the energy of conventional attention modules.

تقدم دراسة جديدة معزز الديناميات الزمنية (TDE) لشبكات الأعصاب النابضة (SNN)، بهدف تحسين قدرتها على نمذجة المعلومات الزمنية في مهام مثل اكتشاف الكائنات. يتضمن TDE مشفر نابض ووحدة توجيه الانتباه، والتي تعمل معًا على توليد محفزات متنوعة وإدارة الاعتماديات الزمنية، مما يعزز القدرة التعبيرية لـ SNN.

Un nuevo estudio presenta el Mejorador de Dinámicas Temporales (TDE) para Redes Neuronales Espiking (SNN), con el objetivo de mejorar su capacidad para modelar información temporal en tareas como la detección de objetos. El TDE incluye un Codificador Espiking y un Módulo de Atención, que juntos generan estímulos diversos y gestionan las dependencias intertemporales, mejorando así el poder expresivo de las SNN.

Une nouvelle étude présente l'Amplificateur de Dynamiques Temporelles (TDE) pour les Réseaux de Neurones à Émission de Spikes (SNN), visant à améliorer leur capacité à modéliser l'information temporelle dans des tâches telles que la détection d'objets. Le TDE comprend un Encodeur à Émission de Spikes et un Module de Gating d'Attention, qui génèrent ensemble des stimuli divers et gèrent les dépendances inter-temporelles, renforçant ainsi le pouvoir expressif des SNN.

A new study introduces the Temporal Dynamics Enhancer (TDE) for Spiking Neural Networks (SNNs), aiming to improve their ability to model temporal information in tasks like object detection. TDE includes a Spiking Encoder and an Attention Gating Module, which together generate diverse stimuli and manage inter-temporal dependencies, enhancing the expressive power of SNNs.

Exploring the Potentials of Spiking Neural Networks for Image Deraining

Was this article worth reading? Share it