Research map / Detection and segmentation / Open-set and few-shot detection
Thread Counting in Plain Weave for Old Paintings Using Regression Deep Learning Models
Antonio Delgado, Juan José Murillo-Fuentes, Laura Alba Carcelén
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 segmentation with direct regression (Reg-VGG) | Inc-Dice thread counting · test set | NMAE (%): 1.51 -> 1.02 | Improved |
| replace segmentation with direct regression (Reg-VGG) | Inc-Dice thread counting · validation set | NMAE (%): 1.12 -> 0.92 | Improved |
| add preprocessing improvements (equalization, variable window, crops, limited rotation) to segmentation | Inc-Dice · validation set | NMAE (%): 1.12 -> 1.04 | Improved |
| add the dataset-generation improvements | Inc-Dice · test set | NMAE (%): 1.61 -> 1.51 | Improved |
| reuse pre-trained segmentation encoder instead of training from scratch | regression model · validation set | NMAE (%) | Hurt |
| add residual connections to inception regression | Reg · validation set | NMAE (%) | No clear effect |
| use VGG-style layout instead of plain inception regression | Reg · validation set | NMAE (%): 0.92 | Improved |
| add histogram equalization to preprocessing (ablation) | Reg-VGG · test set | NMAE | Improved |
| add variable-size filtering window instead of fixed window (ablation) | Reg-VGG · test set | NMAE | Improved |
| add central patches from labeled samples to data augmentation (ablation) | Reg-VGG · test set | NMAE | Improved |
| limit maximum random rotation angle in augmentation (ablation) | Reg-VGG · test set | NMAE | Improved |
| add semi-supervised refinement using FT agreement | Reg-VGG · test set | NMAE (%): 1.01 -> 0.94 | Improved |
Closest papers by meaning
- The Met Dataset: Instance-level Recognition for Artworks (NeurIPS 2021 · similarity 0.83)
- A Deep Learning Approach to Clustering Visual Arts (IJCV 2022 · similarity 0.83)
- TexTile: A Differentiable Metric for Texture Tileability (CVPR 2024 · similarity 0.80)
- APDDv2: Aesthetics of Paintings and Drawings Dataset with Artist Labeled Scores and Comments (NeurIPS 2024 · similarity 0.80)
- Learning to Count without Annotations (CVPR 2024 · similarity 0.80)
- Reconstructing the Image Stitching Pipeline: Integrating Fusion and Rectangling into a Unified Inpainting Model (NeurIPS 2024 · similarity 0.80)
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.
