Research map / Detection and segmentation / CNN features
A Holistically-Guided Decoder for Deep Representation Learning With Applications to Semantic Segmentation and Object Detection
Jianbo Liu, Junjun He, Yuanjie Zheng, Shuai Yi, Xiaogang Wang, Hongsheng Li
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 bilinear/deconv skip-connected decoder with HGD | EfficientFCN · PASCAL Context val | mIoU (%): 49.3 -> 55.3 | Improved |
| add OS=16 and OS=8 maps to codeword-generation fusion (ablation) | EfficientFCN · PASCAL Context val | mIoU (%): 54.8 -> 55.3 | Improved |
| add OS=8 and OS=16 fusion to coefficient estimation (ablation) | EfficientFCN · PASCAL Context val | mIoU (%): 47.9 -> 55.3 | Improved |
| add additive codeword information transfer to coefficient branch (ablation) | EfficientFCN · PASCAL Context val | mIoU (%): 54.2 -> 55.3 | Improved |
| increase number of codewords from 32 to 512 (ablation) | EfficientFCN · PASCAL Context val | mIoU (%): 54.5 -> 55.5 | Improved |
| increase codewords from 512 to 1024 (ablation) | EfficientFCN · PASCAL Context val | mIoU (%): 55.5 -> 55.1 | Hurt |
| add holistic codewords on top of multi-scale fusion only (n=0,c=0) (ablation) | HGD-FPN Faster R-CNN · COCO minival | AP: 37.5 -> 40.0 | Improved |
| add multi-scale fusion only (n=0,c=0) to FPN (ablation) | Faster R-CNN FPN · COCO minival | AP: 36.3 -> 37.5 | Improved |
| increase recurrence k from 1 to 4 (ablation) | HGD-FPN Faster R-CNN · COCO minival | AP: 37.9 -> 40.0 | Improved |
| increase recurrence k from 4 to 5 (ablation) | HGD-FPN Faster R-CNN · COCO minival | AP: 40.0 -> 39.8 | Hurt |
| share parameters across recurrent HGD-FPN stages (ablation) | HGD-FPN Faster R-CNN · COCO minival | AP: 39.6 -> 40.0 | Improved |
| replace FPN with HGD-FPN | RetinaNet · COCO minival | AP: 35.3 -> 37.8 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CARAFE — CARAFE: Content-Aware ReAssembly of FEatures (ICCV 2019) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- EfficientFCN: Holistically-guided Decoding for Semantic Segmentation (ECCV 2020 · similarity 0.95)
- Dense Decoder Shortcut Connections for Single-Pass Semantic Segmentation (CVPR 2018 · similarity 0.90)
- Decoders Matter for Semantic Segmentation: Data-Dependent Decoding Enables Flexible Feature Aggregation (CVPR 2019 · similarity 0.89)
- Dynamic Neural Representational Decoders for High-Resolution Semantic Segmentation (NeurIPS 2021 · similarity 0.89)
- SED: A Simple Encoder-Decoder for Open-Vocabulary Semantic Segmentation (CVPR 2024 · similarity 0.87)
- DECO: Unleashing the Potential of ConvNets for Query-based Detection and Segmentation (ICLR 2025 · similarity 0.87)
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.
