How zhn66776/gpr_yolo works
This repository is a GPR hyperbola-detection dataset and YOLOv5 experiment, rather than a self-contained application implementation. Its documented workflow is to prepare annotated GPR images, train a pretrained YOLOv5s model, evaluate it, and run inference to produce detections and visual comparisons. The tree contains dataset variants and labels, but no model code or training/inference scripts; those workflow links therefore reflect README documentation, not inspected source call wiring.
Main entry points: yolo_data Dataset variants · images Augmented scans · labels YOLO labels · data.yaml Dataset config
Components 13
GPR data
- GPR scans
- Augmented scans
GPR-2/train/images - YOLO labels
GPR-2/train/labels - Dataset variants
yolo_data - Dataset config
GPR-2/data.yaml
Detection workflow
- YOLOv5s
- Model training
Evaluation and output
- Model evaluation
- Trained weights
- Image inference
- Detection outputs
- Visual comparisons
Other components
- GPR analyst
Connections 14
- GPR scans → Augmented scans: augmentations
- Augmented scans → Dataset variants: organizes images
- YOLO labels → Dataset variants: labels images
- Dataset config → Model training: configures data
- Dataset variants → Model training: training data
- YOLOv5s → Model training: initializes model
- Model training → Trained weights: produces weights
- Trained weights → Model evaluation: evaluates model
- Model evaluation → Model training: reports metrics
- Trained weights → Image inference: loads weights
- GPR analyst → Image inference: submits scans
- Image inference → Detection outputs: saves detections
- Detection outputs → Visual comparisons: compared with labels
- GPR analyst → Visual comparisons: reviews