Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Dynamic Conditional Networks for Few-Shot Learning
Fang Zhao, Jian Zhao, Shuicheng Yan, Jiashi Feng
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 |
|---|---|---|---|
| replace plain FC conditional layer with 8-D dynamic conditional layer | ResNet-34 conditional net · Amazon Bin Image identification | accuracy: 74.48 -> 76.81 | Improved |
| replace plain FC conditional layer with 8-D dynamic layer | ResNet-34 conditional net · Amazon Bin Image verification | accuracy: 84.87 -> 85.48 | Improved |
| increase dynamic layer dimension from 8 to 16 (ablation) | DCCN · Amazon Bin Image identification | accuracy: 76.81 -> 75.66 | Hurt |
| increase dynamic layer dimension from 4 to 8 (ablation) | DCCN · Amazon Bin Image identification | accuracy: 76.60 -> 76.81 | Improved |
| replace plain network with 64-D dynamic layer | ResNet-34 multi-modal classifier · MIRFlickr-25K 10k tags | IoU: 0.6516 -> 0.6606 | Improved |
| replace plain network with 64-D dynamic layer | ResNet-34 multi-modal classifier · MIRFlickr-25K 5k tags | IoU: 0.6489 -> 0.6553 | Improved |
| increase dynamic layer dimension from 64 to 128 (ablation) | DCCN · MIRFlickr-25K 10k tags | IoU: 0.6606 -> 0.6516 | Hurt |
| use 16-D dynamic layer instead of fully connected layers | GroundeR-style grounding · Flickr30k Entities | accuracy: 47.81 -> 50.65 | Improved |
| increase dynamic layer dimension from 16 to 32 (ablation) | DCCN · Flickr30k Entities | accuracy: 50.65 -> 50.52 | No clear effect |
| dynamic layer versus GroundeR on People phrases | GroundeR · Flickr30k Entities People | accuracy: 61.00 -> 67.37 | Improved |
| dynamic layer versus GroundeR on Vehicles and Scene phrases | GroundeR · Flickr30k Entities Vehicles/Scene | accuracy: 68.75 -> 56.04; 58.18 -> 54.05 | Hurt |
| use dynamic conditional layers in conditional GAN | plain conditional GAN · MS-Celeb-1M face generation | rank-1 accuracy: 0.457 -> 0.688 | Improved |
Closest papers by meaning
- Learning Instance and Task-Aware Dynamic Kernels for Few-Shot Learning (ECCV 2022 · similarity 0.87)
- CondConv: Conditionally Parameterized Convolutions for Efficient Inference (NeurIPS 2019 · similarity 0.86)
- Conditional Image Generation with PixelCNN Decoders (NeurIPS 2016 · similarity 0.86)
- Dynamic Few-Shot Visual Learning Without Forgetting (CVPR 2018 · similarity 0.86)
- Dynamic Capacity Networks (ICML 2016 · similarity 0.85)
- KernelWarehouse: Rethinking the Design of Dynamic Convolution (ICML 2024 · 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.
