Research · 3 entries

Research that measures.

Model evaluation, efficient deep learning and the reliability of retrieval-augmented systems.

  1. SE
  2. ECA
  3. LCA (proposed)

Publication · arXiv · Jan 2026 · Co-author

Lightweight Channel Attention for Efficient CNNs

An empirical study of SE, ECA and a proposed Lite Channel Attention module across ResNet-18 and MobileNetV2 on CIFAR-10, measuring accuracy, parameters, FLOPs and latency.

ResNet-18 accuracy
94.68%
MobileNetV2 accuracy
93.10%