LayoutMatch/inference.py

import faiss
import numpy as np
import torch
from models.rotation_cnn import RotationInvariantNet, get_rotational_features
from data_units import layout_to_tensor, tile_layout

def main():
    # 配置参数（需根据实际调整）
    block_size = 64           # 分块尺寸
    target_module_path = "target.png"
    large_layout_path = "layout_large.png"

    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model = RotationInvariantNet().to(device)
    model.load_state_dict(torch.load("rotation_cnn.pth"))
    model.eval()

    # 预处理目标模块与大版图
    target_tensor = layout_to_tensor(target_module_path, (block_size, block_size))
    target_feat = get_rotational_features(model, torch.tensor(target_tensor).to(device))

    large_layout = layout_to_tensor(large_layout_path)
    tiles = tile_layout(large_layout)

    # 构建特征索引（使用Faiss加速）
    index = faiss.IndexFlatL2(64)  # 特征维度由模型决定
    features_db = []
    for (x, y, tile) in tiles:
        feat = get_rotational_features(model, torch.tensor(tile).to(device))
        features_db.append(feat)
    index.add(np.stack(features_db))

    # 检索相似区域
    D, I = index.search(target_feat[np.newaxis, :], k=10)

    for idx in I[0]:
        x, y, _ = tiles[idx]

        # 计算最佳匹配角度的显式计算
        min_angle, min_dist = 90, float('inf')
        target_vec = target_feat
        feat = features_db[idx]
        for a in [0, 1, 2, 3]:  # 代表0°、90°、180°、270°
            rotated_feat = np.rot90(feat.reshape(block_size, block_size), k=a)
            dist = np.linalg.norm(target_vec - rotated_feat.flatten())
            if dist < min_dist:
                min_dist, min_angle = dist, a * 90

        print(f"坐标({x},{y}), 最佳旋转方向{min_angle}度，距离: {min_dist}")
if __name__ == "__main__":
    main()