Research map / Architectures and efficiency / CNN design
Streaming Convolutional Neural Networks for End-to-End Learning With Multi-Megapixel Images
Hans Pinckaers, Bram van Ginneken, Geert J. S. Litjens
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| increase input resolution from 1024x1024 to 4096x4096 (ablation) | Streaming CNN · TUPAC16 test set | Spearman rho: 0.485 -> 0.536 | Improved |
| increase input from 2048x2048 to 4096x4096 with deeper network (ablation) | Streaming CNN · TUPAC16 test set | Spearman rho: 0.491 -> 0.570 | Improved |
| increase input from 4096x4096 to 8192x8192 (ablation) | Streaming CNN · TUPAC16 test set | Spearman rho: 0.570 -> 0.560 | No clear effect |
| increase input from 4096x4096 to 8192x8192 (ablation) | Streaming CNN · TUPAC16 tuning set | Spearman rho: 0.644 -> 0.692 | Improved |
| increase input resolution from 2048x2048 to 8192x8192 (ablation) | Streaming CNN · CAMELYON17 macro-metastasis | AUC: 0.515 -> 0.827 | Improved |
| increase input resolution from 2048x2048 to 8192x8192 (ablation) | Streaming CNN · CAMELYON17 negative | AUC: 0.580 -> 0.706 | Improved |
| increase input resolution from 2048x2048 to 8192x8192 (ablation) | Streaming CNN · CAMELYON17 isolated tumor cells | AUC: 0.450 -> 0.463 | No clear effect |
| increase input resolution from 2048x2048 to 8192x8192 (ablation) | Streaming CNN · CAMELYON17 micro-metastasis | AUC: 0.689 -> 0.709 | Improved |
| stream with 16 tiles instead of full image | conv layers · 1024x1024 input | peak memory: 1957 MB -> 202 MB | Improved |
| stream with 16 tiles instead of full image | conv layers · 1024x1024 input | backward time: 45 ms -> 147 ms | Hurt |
| train with streaming instead of conventional training | small CNN · ImageNette | training loss | No clear effect |
Closest papers by meaning
- Efficient Classification of Very Large Images With Tiny Objects (CVPR 2022 · similarity 0.89)
- A Dataset Generation Framework for Evaluating Megapixel Image Classifiers & Their Explanations (ECCV 2022 · similarity 0.87)
- Inference, Learning and Attention Mechanisms that Exploit and Preserve Sparsity in CNNs (IJCV 2020 · similarity 0.86)
- Scale-Space Hypernetworks for Efficient Biomedical Image Analysis (NeurIPS 2023 · similarity 0.86)
- Deep Tensor Convolution on Multicores (ICML 2017 · similarity 0.85)
- NaFV-Net: An Adversarial Four-view Network for Mammogram Classification (AAAI 2025 · similarity 0.84)
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.
