Add training API for vector data rasterization

[Copilot]

Implements a Python library for rasterizing vector data (points, lines, polygons) with a direct training API for machine learning workflows.

Core Components

• Rasterizer - Converts vector geometries to raster grids with rasterize_points(), rasterize_lines(), rasterize_polygon(), and batch_rasterize()
• Trainer - High-level training API with three usage modes:
  • prepare_training_data() - Rasterize and normalize for any ML framework
  • train() - Train models directly with prepared data
  • train_from_raw_data() - One-step pipeline from raw vectors to trained model

Usage

from rasterizer import Trainer
import numpy as np

trainer = Trainer(width=256, height=256)

# Option 1: Prepare data for any framework (TensorFlow/PyTorch/sklearn)
data = [{'type': 'points', 'data': np.array([[10, 20], [30, 40]]), 'radius': 2.0}]
X_train, y_train = trainer.prepare_training_data(data, labels)

# Option 2: One-step training
trainer.train_from_raw_data(data, labels, model=your_model, epochs=10)

Implementation Details

• Handles degenerate cases (identical coordinates, collinear vertices) by centering at image midpoint
• Framework-agnostic: accepts any model with fit() method
• Normalization with proper zero-division handling using EPSILON = 1e-8
• Returns shape (batch, height, width, 1) for CNN compatibility

Testing

20 unit tests covering core functionality and edge cases, bilingual documentation with examples.

Original prompt

这个库是否有提供一个api供我直接使用训练我的数据？

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.