Research map / Architectures and efficiency / Attention and transformers
DIANet: Dense-and-Implicit Attention Network
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| share a DIA-LSTM attention module across layers | ResNet164 · CIFAR100 | top1 accuracy: 75.03 -> 76.67 | Helped |
| share a DIA-LSTM attention module across layers | ResNet50 · ImageNet 2012 | top1 accuracy: 76.61 -> 77.24 | Helped |
| share SE module across layers instead of using individual SE modules | ResNet164 · CIFAR100 | top1 accuracy: 75.03 -> 75.74 | Helped |
| increase reduction ratio from 1 to 16 (ablation) | DIANet ResNet164 · CIFAR100 | top1 accuracy: 76.88 -> 76.50 | Mixed |
| replace sigmoid output activation with tanh (ablation) | DIANet ResNet164 · CIFAR100 | top1 accuracy: 76.67 -> 75.24 | Hurt |
| replace sigmoid output activation with ReLU (ablation) | DIANet ResNet164 · CIFAR100 | top1 accuracy: 76.67 -> 74.62 | Hurt |
| add two stacked LSTM cells (ablation) | DIANet ResNet164 with sigmoid output · CIFAR100 | top1 accuracy: 76.67 -> 75.20 | Hurt |
| add two stacked LSTM cells (ablation) | DIANet ResNet164 with tanh output · CIFAR100 | top1 accuracy: 75.24 -> 76.47 | Helped |
| add DIA-LSTM in stage 1 (ablation) | DIANet ResNet164 · CIFAR100 | top1 accuracy: 76.27 -> 76.67 | Helped |
| add DIA-LSTM in stage 2 (ablation) | DIANet ResNet164 · CIFAR100 | top1 accuracy: 76.25 -> 76.67 | Helped |
| add DIA-LSTM in stage 3 (ablation) | DIANet ResNet164 · CIFAR100 | top1 accuracy: 75.40 -> 76.67 | Helped |
| use DIANet rather than SENet | ResNet164 without data augmentation · CIFAR10 | test accuracy: 88.30 -> 89.25 | Helped |
Closest papers by meaning
- Residual Attention Network for Image Classification (CVPR 2017 · similarity 0.89)
- SimAM: A Simple, Parameter-Free Attention Module for Convolutional Neural Networks (ICML 2021 · similarity 0.87)
- Learn to Pay Attention (ICLR 2018 · similarity 0.87)
- Less Is More Important: An Attention Module Guided by Probability Density Function for Convolutional Neural Networks (AAAI 2023 · similarity 0.87)
- EAN: An Efficient Attention Module Guided by Normalization for Deep Neural Networks (AAAI 2024 · similarity 0.86)
- LSNet: See Large, Focus Small (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.