Research map / Detection and segmentation / CNN features
PRUE: A Practical Recipe for Field Boundary Segmentation at Scale
Gedeon Muhawenayo, Caleb Robinson, Subash Khanal, Zhanpei Fang, Isaac Corley, Alexander Wollam, Tianyi Gao, Leonard Strnad, Ryan Avery, Lyndon Estes, Ana Tárano, Nathan Jacobs and 1 more
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 the loss with log-cosh Dice (ablation) | U-Net EfficientNet-B3 on FTW · FTW test | Pixel IoU: 0.68 -> 0.77 | Improved |
| set boundary class weight to 0.75 (ablation) | U-Net EfficientNet-B3 on FTW · FTW test | Object F1: 0.39 -> 0.42 | Improved |
| use EfficientNet-B7 encoder instead of B3 (ablation) | U-Net on FTW · FTW test | Object F1: 0.39 -> 0.42 | Improved |
| add brightness and resize augmentations (ablation) | U-Net EfficientNet-B3 on FTW · FTW test | Scale sensitivity F1|delta|: 0.15 -> 0.00 | Improved |
| add brightness and resize augmentations (ablation) | U-Net EfficientNet-B3 on FTW · FTW test | Object F1: 0.39 -> 0.38 | No clear effect |
| add channel shuffle augmentation (ablation) | U-Net EfficientNet-B3 on FTW · FTW test | Input-order sensitivity F1|delta|: 0.07 -> 0.00 | Improved |
| replace U-Net with FCSiam (ablation) | EfficientNet-B3 on FTW · FTW test | Object F1: 0.39 -> 0.40 | No clear effect |
| combine B7 encoder, log-cosh Dice, weight 0.75, brightness+resize and channel shuffle (ablation) | FTW baseline U-Net · FTW test | Object F1: 0.38 -> 0.47 | Improved |
| use frozen geospatial foundation model encoders with a trained decoder instead of U-Net | FTW field boundary segmentation · FTW test | Pixel IoU: 0.76 -> 0.67 (best GFM, Clay) | Hurt |
| use instance segmentation (zero-shot SAM or Delineate Anything) instead of fine-tuned U-Net | FTW field boundary segmentation · FTW test | Pixel IoU: 0.70 -> 0.32 (SAM zero-shot) | Hurt |
| scale up from U-Net B3 to B7 combination | inference throughput · FTW test | Throughput (km2/s): 623.28 -> 306.94 | Hurt |
Closest papers by meaning
- Fields of The World: A Machine Learning Benchmark Dataset for Global Agricultural Field Boundary Segmentation (AAAI 2025 · similarity 0.93)
- Can Large Multimodal Models Understand Agricultural Scenes? Benchmarking with AgroMind (NeurIPS 2025 · similarity 0.88)
- GTPBD: A Fine-Grained Global Terraced Parcel and Boundary Dataset (NeurIPS 2025 · similarity 0.87)
- Exact: Exploring Space-Time Perceptive Clues for Weakly Supervised Satellite Image Time Series Semantic Segmentation (CVPR 2025 · similarity 0.87)
- Towards Open-Vocabulary Remote Sensing Image Semantic Segmentation (AAAI 2025 · similarity 0.86)
- Combining Deep Learning and Street View Imagery to Map Smallholder Crop Types (AAAI 2024 · similarity 0.86)
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.
