Research map / Detection and segmentation / CNN features
The Devil is in the Decoder: Classification, Regression and GANs
Zbigniew Wojna, Vittorio Ferrari, Sergio Guadarrama, Nathan Silberman, Liang-Chieh Chen, Alireza Fathi, Jasper Uijlings
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 |
|---|---|---|---|
| add residual-like connections to decoders (ablation) | decoders across tasks · seven pixel-wise tasks | task metrics | Improved |
| add residual-like connections (ablation) | stacked hourglass decoder · Human keypoints estimation | PCKh | No clear effect |
| add residual connections to transposed convolution decoder (ablation) | Transposed · Depth regression artifacts, x-axis | artifact rate: 27.6% -> 23.1% | Improved |
| add skip connections to transposed convolution decoder (ablation) | Transposed · Depth regression artifacts, x-axis | artifact rate: 27.6% -> 30.6% | Hurt |
| add skip connections to decomposed transposed decoder (ablation) | Dec. Transposed · Depth regression artifacts, x-axis | artifact rate: 27.0% -> 33.4% | Hurt |
| add skip connections to bilinear upsampling + conv (ablation) | Bilinear Ups. + Conv. · Depth regression artifacts, x-axis | artifact rate: 14.3% -> 19.5% | Hurt |
| use bilinear upsampling + conv instead of transposed convolution (ablation) | decoder for depth regression · Depth regression artifacts, x-axis | artifact rate: 27.6% -> 14.3% | Improved |
| use bicubic instead of nearest neighbour upsampling (ablation) | Bilinear Ups. + Conv, res+skip · Human keypoints estimation | PCK/0.50: 0.846 -> 0.855 | No clear effect |
| use bilinear instead of nearest neighbour upsampling (ablation) | Bilinear Ups. + Conv, res+skip · Human keypoints estimation | PCK/0.50: 0.846 -> 0.854 | No clear effect |
| use bilinear instead of nearest neighbour upsampling (ablation) | Bilinear Ups. + Conv, res · GANs on Cifar 10 | FID: 28.0 -> 29.8 | Hurt |
| use bicubic instead of nearest neighbour upsampling (ablation) | Bilinear Ups. + Conv, res · GANs on Cifar 10 | FID: 28.0 -> 28.1 | No clear effect |
| use depth-to-space upsampling without residual connections (ablation) | decoder · Semantic segmentation | mIoU | Improved |
Closest papers by meaning
- Decoders Matter for Semantic Segmentation: Data-Dependent Decoding Enables Flexible Feature Aggregation (CVPR 2019 · similarity 0.88)
- Deep Decoder: Concise Image Representations from Untrained Non-convolutional Networks (ICLR 2019 · similarity 0.87)
- Win-Win: Training High-Resolution Vision Transformers from Two Windows (ICLR 2024 · similarity 0.85)
- Large Vision-Language Models Get Lost in Attention (ICML 2026 · similarity 0.85)
- Rethinking Decoders for Transformer-based Semantic Segmentation: A Compression Perspective (NeurIPS 2024 · similarity 0.85)
- Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis (NeurIPS 2026 · similarity 0.85)
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.
