A HaloNet is a self-attention based model for efficient image classification. It relies on a local self-attention architecture that efficiently maps to existing hardware with haloing. The formulation breaks translational equivariance, but the authors observe that it improves throughput and accuracies over the centered local self-attention used in regular self-attention. The approach also utilises a strided self-attentive downsampling operation for multi-scale feature extraction.
Source: Scaling Local Self-Attention for Parameter Efficient Visual BackbonesPaper | Code | Results | Date | Stars |
---|
Task | Papers | Share |
---|---|---|
Anomaly Detection | 1 | 16.67% |
Unsupervised Anomaly Detection | 1 | 16.67% |
Image Classification | 1 | 16.67% |
Instance Segmentation | 1 | 16.67% |
Object Detection | 1 | 16.67% |
Semantic Segmentation | 1 | 16.67% |
Component | Type |
|
---|---|---|
🤖 No Components Found | You can add them if they exist; e.g. Mask R-CNN uses RoIAlign |