Research map / Architectures and efficiency / CNN design
Dictionary Learning Inspired Deep Network for Scene Recognition
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 |
|---|---|---|---|
| replace FCL+ReLU with dictionary learning layers (both FC7, FC8) (ablation) | AlexNet CNN-DL · MIT Indoor 67 | accuracy (%): 61.80 -> 65.19 | Helped |
| replace one FCL with DLL (FC8) (ablation) | AlexNet CNN-DL · MIT Indoor 67 | accuracy (%): 61.80 -> 64.63 | Helped |
| use DLL instead of LISTA layers (ablation) | AlexNet CNN-DL · MIT Indoor 67 | accuracy (%): 63.90 -> 65.19 | Helped |
| use two recurrent units instead of one (ablation) | AlexNet CNN-DL · MIT Indoor 67 | accuracy (%): 63.21 -> 64.63 | Helped |
| increase recurrent units from 2 to 8 (ablation) | AlexNet CNN-DL · MIT Indoor 67 | accuracy (%): 64.63 -> 63.08 | Hurt |
| add label discriminative regressor (L2) to loss (ablation) | AlexNet CNN-DL · MIT Indoor 67 | accuracy (%): 65.19 -> 66.11 | Helped |
| add overfitting constraint on B (Mahalanobis/Euclidean balance) (ablation) | AlexNet CNN-DL · MIT Indoor 67 | accuracy (%): 66.11 -> 66.60 | Helped |
| increase dictionary size per class beyond 40 (ablation) | AlexNet CNN-DL · MIT Indoor 67 | accuracy (%) | No clear effect |
| replace FCL+ReLU with DLL on VGG ImageNet-pretrained | VGG CNN-DL · MIT Indoor 67 | accuracy (%): 71.11 -> 78.33 | Helped |
| replace FCL+ReLU with DLL on AlexNet ImageNet-pretrained | AlexNet CNN-DL · SUN 397 | accuracy (%): 46.33 -> 51.67 | Helped |
Closest papers by meaning
- Dictionary Pair Classifier Driven Convolutional Neural Networks for Object Detection (CVPR 2016 · similarity 0.86)
- Deep Semantic Dictionary Learning for Multi-label Image Classification (AAAI 2021 · similarity 0.84)
- Neural Implicit Dictionary Learning via Mixture-of-Expert Training (ICML 2022 · similarity 0.84)
- Depth CNNs for RGB-D Scene Recognition: Learning from Scratch Better than Transferring from RGB-CNNs (AAAI 2017 · similarity 0.84)
- Learning Dictionary for Visual Attention (NeurIPS 2023 · similarity 0.83)
- Deep Convolutional Dictionary Learning for Image Denoising (CVPR 2021 · similarity 0.83)
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.