EeveeDark: A Binary Neural Framework for Low-Light Video Enhancement via Event-Guided Sensor-Level Fusion

2026-07-08 · Source: cs.CV updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Computer Vision · Depth: Expert, extended

Summary

EeveeDark is a novel binary neural framework designed for low-light video enhancement, integrating the spatial richness of sensor-level RAW data with the temporal precision of event streams. Accepted on February 10, 2026, this framework utilizes a Binary Neural Network (BNN) architecture to significantly reduce computational overhead by quantizing weights and activations. It features modality-specific binary encoders for RAW frames and event data, a lightweight fusion block, and an event-guided skip gating mechanism for dynamic spatiotemporal refinement. Experiments on synthetic (LLRVD, SDSD) and real-world (HUE, SDE) datasets demonstrate EeveeDark's superior performance over prior BNN-based methods like BBCU and BRVE, achieving 37.51 PSNR / 0.039 STRRED on LLRVD. It also offers a compelling performance-efficiency trade-off against full-precision models, operating at 1.66G FLOPs and 0.35M parameters, with an estimated latency of ~588 ms and 0.78 mJ per frame. EeveeDark enhances downstream robotic perception tasks, including object detection (mAP 0.73) and visual SLAM.

Key takeaway

For Robotics Engineers deploying low-light video enhancement on resource-constrained platforms, EeveeDark presents a compelling solution. Its Binary Neural Network architecture and RAW-event fusion deliver high-quality, temporally consistent video at significantly lower computational cost than full-precision models. You should evaluate EeveeDark to achieve robust perception in challenging illumination conditions, improving downstream tasks like object detection and visual SLAM without excessive hardware demands.

Key insights

Fusing RAW data and event streams with Binary Neural Networks enables efficient, high-quality low-light video enhancement.

Principles

• Binary Neural Networks offer extreme efficiency.
• RAW data preserves critical sensor-level information.
• Event streams provide high temporal resolution.

Method

EeveeDark preprocesses RAW frames and event streams, encodes them with modality-specific binary encoders, fuses features, and refines them via a Shift Encoder with Event-Guided Skip Gates, then decodes and reconstructs.

In practice

• Deploy BNNs for resource-constrained robotic vision.
• Integrate RAW and event data for robust low-light perception.
• Enhance video early in the imaging pipeline for fidelity.

Topics

• EeveeDark
• Low-Light Video Enhancement
• Binary Neural Networks
• Event Cameras
• RAW Data Processing
• Robotic Perception

Code references

• MichaelGrupp/evo

Best for: Computer Vision Engineer, Research Scientist, AI Scientist, Machine Learning Engineer, Robotics Engineer

Related on AIssential

• NERVE: A Neuromorphic Vision and Radar Ensemble for Multi-Sensor Fusion Research — NERVE is a multi-sensor dataset bridging neuromorphic vision and radar for robust perception research.
• Neural Events: Discrete Asynchronous Autoencoders for Event-Based Vision — The framework re-tokenizes high-resolution event streams into compressed, informative "neural events" using discrete learnable codes.
• Multinex: An ultra lightweight AI model advancing low light image enhancement — Multinex uses classical color vision and Retinex principles for efficient, compact low-light image enhancement.
• SNNF: An SNN-based Near-Sensor Noise Filter for Dynamic Vision Sensors — SNNF efficiently filters DVS noise using a compact binary representation and a hardware-optimized Spiking Neural Network.
• NeuroLip: An Event-driven Spatiotemporal Learning Framework for Cross-Scene Lip-Motion-based Visual Speaker Recognition — Event cameras and specialized processing can enable robust lip-motion-based speaker recognition across diverse scenes.
• DarkDriving: A Real-World Day and Night Aligned Dataset for Autonomous Driving in the Dark Environment — DarkDriving offers the first real-world dataset with precisely aligned day-night image pairs for autonomous driving.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CV updates on arXiv.org.