REVIEW 6 cited by
Spike Calibration: Fast and Accurate Conversion of Spiking Neural Network for Object Detection and Segmentation
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
abstract
Spiking neural network (SNN) has been attached to great importance due to the properties of high biological plausibility and low energy consumption on neuromorphic hardware. As an efficient method to obtain deep SNN, the conversion method has exhibited high performance on various large-scale datasets. However, it typically suffers from severe performance degradation and high time delays. In particular, most of the previous work focuses on simple classification tasks while ignoring the precise approximation to ANN output. In this paper, we first theoretically analyze the conversion errors and derive the harmful effects of time-varying extremes on synaptic currents. We propose the Spike Calibration (SpiCalib) to eliminate the damage of discrete spikes to the output distribution and modify the LIPooling to allow conversion of the arbitrary MaxPooling layer losslessly. Moreover, Bayesian optimization for optimal normalization parameters is proposed to avoid empirical settings. The experimental results demonstrate the state-of-the-art performance on classification, object detection, and segmentation tasks. To the best of our knowledge, this is the first time to obtain SNN comparable to ANN on these tasks simultaneously. Moreover, we only need 1/50 inference time of the previous work on the detection task and can achieve the same performance under 0.492$\times$ energy consumption of ANN on the segmentation task.
Forward citations
Cited by 6 Pith papers
-
Ultra-Low-Latency Spiking Neural Networks with Temporal-Dependent Integrate-and-Fire Neuron Model for Objects Detection
Binary temporal coding in a new spiking neuron (tdIF) plus delay-spike achieves strong ANN-to-SNN conversion results on detection and lane tasks, but the reported five-step latency ignores the depth-multiplied pipeline delay.
-
Integer Binary-Range Alignment Neuron for Spiking Neural Networks
A binary-encoded integer spiking neuron with range alignment matches or beats ANN accuracy on ImageNet, COCO, and CIFAR100 at lower energy cost.
-
Quantized Spike-driven Transformer
A 4-bit quantized spike-driven transformer with multi-bit training and binary inference achieves 80.3% ImageNet accuracy with 6.8M parameters.
-
CREST: An Efficient Conjointly-trained Spike-driven Framework for Event-based Object Detection Exploiting Spatiotemporal Dynamics
CREST combines a conjoint learning rule, a multi-scale spatiotemporal event integrator, and a spatiotemporal IoU loss to improve accuracy and estimated energy efficiency of event-based object detection with spiking ne...
-
Spike2Former: Efficient Spiking Transformer for High-performance Image Segmentation
A spiking Mask2Former with normalized integer neurons achieves state-of-the-art SNN performance on ADE20K, CityScapes, and Pascal VOC, with claimed large energy savings.
-
Scaling Spike-driven Transformer with Efficient Spike Firing Approximation Training
A spike-driven Transformer trained with integer activations reaches 86.2% top-1 on ImageNet, the highest reported accuracy for a directly trained spiking network at this scale.
Discussion (0). Continue with ORCID to comment.