Research map / Image generation / Text-to-image diffusion
Annotation-Free Human Sketch Quality Assessment
Lan Yang, Kaiyue Pang, Honggang Zhang, Yi-Zhe Song
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 soft binning (ablation) | GACL Ours-Cos · QuickDraw human study | human agreement: 72.35% -> 78.13% | Improved |
| use quality-aware GACL training instead of plain BiLSTM | BiLSTM · QuickDraw test split | top-1 accuracy: 79.87% -> 82.52% | Improved |
| GACL Ours-Cos vs strongest baseline SketchAA | GACL · QuickDraw test split | top-1 accuracy: 81.51% -> 82.52% | Improved |
| swap BiLSTM for Transformer point-cloud encoder (ablation) | GACL · QuickDraw human study | human agreement: 78.13% -> 77.87% | No clear effect |
| swap BiLSTM for ConvNeXt raster encoder (ablation) | GACL · QuickDraw human study | human agreement: 78.13% -> 77.44% | No clear effect |
| add RHS stroke-removal conditioning to q scorer (ablation) | q scorer · malicious sketch attribution | F1: 61.25% | Improved |
| use N=100 bins instead of 50 on LIVE (ablation) | GACL IQA · LIVE | SROCC: 0.928 -> 0.958 | Improved |
| use N=100 bins instead of 200 on TID2013 (ablation) | GACL IQA · TID2013 | SROCC: 0.914 -> 0.927 | Improved |
| use N=50 bins instead of 100 on TID2013 (ablation) | GACL IQA · TID2013 | SROCC: 0.927 -> 0.940 | Improved |
| use N=200 bins instead of 100 on KONIQ-10K (ablation) | GACL IQA · KONIQ-10K | SROCC: 0.922 -> 0.926 | No clear effect |
| apply GACL IQA (ResNet50) vs TReS on small LIVE dataset | GACL IQA · LIVE | SROCC: 0.969 -> 0.958 | Hurt |
| apply GACL IQA (ResNet50) vs TReS on TID2013 | GACL IQA · TID2013 | SROCC: 0.863 -> 0.927 | Improved |
Closest papers by meaning
- Finding Badly Drawn Bunnies (CVPR 2022 · similarity 0.93)
- Do Generalised Classifiers really work on Human Drawn Sketches? (ECCV 2024 · similarity 0.87)
- How to Handle Sketch-Abstraction in Sketch-Based Image Retrieval? (CVPR 2024 · similarity 0.84)
- Open Vocabulary Semantic Scene Sketch Understanding (CVPR 2024 · similarity 0.83)
- One-Shot Reference-based Structure-Aware Image to Sketch Synthesis (AAAI 2025 · similarity 0.83)
- Doomed to Re-Annotate, Forever: The ImageNet Story (NeurIPS 2026 · 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.
