Brain-Inspired Computing: A Systematic Survey and Future Trends

Proceedings of the IEEE, Journal Year: 2024, Volume and Issue: 112(6), P. 544 - 584

Published: June 1, 2024

Language: Английский

---

SpikingResformer: Bridging ResNet and Vision Transformer in Spiking Neural Networks

2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Journal Year: 2024, Volume and Issue: 34, P. 5610 - 5619

Published: June 16, 2024

Language: Английский

---

Integer-Valued Training and Spike-Driven Inference Spiking Neural Network for High-Performance and Energy-Efficient Object Detection

Lecture notes in computer science, Journal Year: 2024, Volume and Issue: unknown, P. 253 - 272

Published: Nov. 22, 2024

Language: Английский

---

Biologically Inspired Spatial–Temporal Perceiving Strategies for Spiking Neural Network

Biomimetics, Journal Year: 2025, Volume and Issue: 10(1), P. 48 - 48

Published: Jan. 14, 2025

A future unmanned system needs the ability to perceive, decide and control in an open dynamic environment. In order fulfill this requirement, it construct a method with universal environmental perception ability. Moreover, perceptual process be interpretable understandable, so that interactions between systems humans can unimpeded. However, current mainstream DNN (deep learning neural network)-based AI (artificial intelligence) is 'black box'. We cannot interpret or understand how decision made by these AIs. An SNN (spiking network), which more similar biological brain than DNN, has potential implement understandable AI. work, we propose neuron group-based structural for better capture spatial temporal information from external environment, time-slicing scheme of responses generated SNN. Results show our indeed helps enhance environment SNN, possesses certain degree robustness, enhancing build future.

Language: Английский

---

Spiking Neural Network Target Detection Method Based on Efficient Deep Feature Extraction

Computer Science and Application, Journal Year: 2025, Volume and Issue: 15(01), P. 187 - 198

Published: Jan. 1, 2025

Language: Английский

---

Efficient Spiking Neural Network for RGB–Event Fusion-Based Object Detection

Electronics, Journal Year: 2025, Volume and Issue: 14(6), P. 1105 - 1105

Published: March 11, 2025

Robust object detection in challenging scenarios remains a critical challenge for autonomous driving systems. Inspired by human visual perception, integrating the complementary modalities of RGB frames and event streams presents promising approach to achieving robust detection. However, existing multimodal detectors achieve superior performance at cost significant computational power consumption. To address this challenge, we propose novel spiking RGB–event fusion-based network (SFDNet), fully detector capable both low-power high-performance Specifically, first introduce Leaky Integrate-and-Multi-Fire (LIMF) neuron model, which combines soft hard reset mechanisms enhance feature representation SNNs. We then develop multi-scale hierarchical residual attention lightweight aggregation module efficient dual-modality extraction fusion. Experimental results on two public datasets demonstrate that our SFDNet achieves state-of-the-art with remarkably low The scenarios, such as motion blur low-light conditions, highlights robustness effectiveness SFDNet, significantly advancing applicability SNNs real-world tasks.

Language: Английский

---

Research on target detection for autonomous driving based on ECS-spiking neural networks

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: April 21, 2025

Abstract In response to the increasing demands for improved model performance and reduced energy consumption in object detection tasks relevant autonomous driving, this research presents an advanced YOLO model, designated as ECSLIF-YOLO, which is based on Leaky Integrate-and-Fire with Extracellular Space (ECS-LIF) framework. The primary aim of tackle issues associated high traditional artificial neural networks (ANNs) suboptimal existing spiking (SNNs). Empirical findings demonstrate that ECSLIF-YOLO achieves a peak mean Average Precision (mAP) 0.917 BDD100K KITTI datasets, thereby aligning accuracy levels conventional ANNs while exceeding current direct-training SNN approaches without incurring additional costs. These suggest particularly well-suited assist development efficient reliable systems driving.

Language: Английский

---

CPT-SNN: A spiking neural network that can combine the previous timestep

Neurocomputing, Journal Year: 2025, Volume and Issue: unknown, P. 130253 - 130253

Published: April 1, 2025

Language: Английский

---

EventAugment: Learning Augmentation Policies from Asynchronous Event-based Data

IEEE Transactions on Cognitive and Developmental Systems, Journal Year: 2024, Volume and Issue: 16(4), P. 1521 - 1532

Published: March 22, 2024

Data augmentation is an effective way to overcome the over-fitting problem of deep learning models. However, most existing studies on data work frame-like (e.g., images), and few tackles with event-based data. Event-based are different from data, rendering techniques designed for unsuitable This deals object detection classification, which important self-driving, robot manipulation. Specifically, we introduce EventAugment, a new method augment asynchronous by automatically policies. We first identify 13 types operations augmenting Next, formulate finding optimal policies as hyperparameter optimization problem. To tackle this problem, propose random search-based framework. Finally, evaluate proposed six public datasets including N-Caltech101, N-Cars, ST-MNIST, N-MNIST, DVSGesture DDD17. Experimental results demonstrate that EventAugment exhibits substantial performance improvements both neural network-based spiking models, gains up approximately 4%. Notably, outperform state-of-the-art methods in terms overall performance.

Language: Английский

---

MVT: Multi-Vision Transformer for Event-Based Small Target Detection

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(9), P. 1641 - 1641

Published: May 4, 2024

Object detection in remote sensing plays a crucial role various ground identification tasks. However, due to the limited feature information contained within small targets, which are more susceptible being buried by complex backgrounds, especially extreme environments (e.g., low-light, motion-blur scenes). Meanwhile, event cameras offer unique paradigm with high temporal resolution and wide dynamic range for object detection. These advantages enable without intensity of light, perform better challenging conditions compared traditional cameras. In this work, we introduce Multi-Vision Transformer (MVT), comprises three efficiently designed components: downsampling module, Channel Spatial Attention (CSA) Global (GSA) module. This architecture simultaneously considers short-term long-term dependencies semantic information, resulting improved performance Additionally, propose Cross Deformable (CDA), progressively fuses high-level low-level features instead considering all scales at each layer, thereby reducing computational complexity multi-scale features. Nevertheless, scarcity camera datasets, provide Event Detection (EOD) dataset, is first dataset that includes scenarios specifically introduced using Moreover, conducted experiments on EOD two typical unmanned aerial vehicle datasets (VisDrone2019 UAVDT Dataset). The comprehensive results demonstrate proposed MVT-Net achieves promising competitive performance.

Language: Английский

---