Spaces:
Paused
Paused
Download modules/vision/aligner.py from SamilG2/BACKEND-CheckScan: direct link, hf CLI and curl.
- Browser
- Download file 3.54 kB
-
https://huggingface.co/spaces/SamilG2/BACKEND-CheckScan/resolve/main/modules/vision/aligner.py
- Command line
-
hf download hf://spaces/SamilG2/BACKEND-CheckScan/modules/vision/aligner.py
-
curl -L -o aligner.py https://huggingface.co/spaces/SamilG2/BACKEND-CheckScan/resolve/main/modules/vision/aligner.py
3.54 kB
| import cv2 | |
| import numpy as np | |
| def _order_points(pts): | |
| rect = np.zeros((4, 2), dtype='float32') | |
| s = pts.sum(axis=1) | |
| rect[0] = pts[np.argmin(s)] | |
| rect[2] = pts[np.argmax(s)] | |
| d = np.diff(pts, axis=1) | |
| rect[1] = pts[np.argmin(d)] | |
| rect[3] = pts[np.argmax(d)] | |
| return rect | |
| def _apply_perspective(img, pts): | |
| rect = _order_points(pts.astype('float32')) | |
| width_top = np.linalg.norm(rect[1] - rect[0]) | |
| width_bot = np.linalg.norm(rect[2] - rect[3]) | |
| height_left = np.linalg.norm(rect[3] - rect[0]) | |
| height_right = np.linalg.norm(rect[2] - rect[1]) | |
| max_width = int(max(width_top, width_bot)) | |
| max_height = int(max(height_left, height_right)) | |
| scale = 800 / max_width | |
| target_w = 800 | |
| target_h = max(600, min(int(max_height * scale), 1400)) | |
| dst = np.array([[0, 0], [target_w, 0], [target_w, target_h], [0, target_h]], dtype='float32') | |
| M = cv2.getPerspectiveTransform(rect, dst) | |
| return cv2.warpPerspective(img, M, (target_w, target_h)) | |
| def _method1_paper_edges(img): | |
| gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) | |
| blurred = cv2.GaussianBlur(gray, (7, 7), 0) | |
| edges = cv2.Canny(blurred, 30, 100) | |
| edges = cv2.dilate(edges, None, iterations=2) | |
| contours, _ = cv2.findContours(edges, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) | |
| if not contours: | |
| return None | |
| img_area = img.shape[0] * img.shape[1] | |
| for cnt in sorted(contours, key=cv2.contourArea, reverse=True)[:5]: | |
| if cv2.contourArea(cnt) < img_area * 0.2: | |
| continue | |
| epsilon = 0.02 * cv2.arcLength(cnt, True) | |
| approx = cv2.approxPolyDP(cnt, epsilon, True) | |
| if len(approx) == 4: | |
| return _apply_perspective(img, approx.reshape(4, 2)) | |
| return None | |
| def _method2_grid_lines(img): | |
| gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) | |
| blurred = cv2.GaussianBlur(gray, (5, 5), 0) | |
| binary = cv2.adaptiveThreshold(blurred, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, | |
| cv2.THRESH_BINARY_INV, 15, 10) | |
| h, w = binary.shape | |
| h_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (w // 8, 1)) | |
| h_lines = cv2.morphologyEx(binary, cv2.MORPH_OPEN, h_kernel, iterations=2) | |
| v_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (1, h // 8)) | |
| v_lines = cv2.morphologyEx(binary, cv2.MORPH_OPEN, v_kernel, iterations=2) | |
| grid = cv2.dilate(cv2.add(h_lines, v_lines), None, iterations=3) | |
| contours, _ = cv2.findContours(grid, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) | |
| if not contours: | |
| return None | |
| largest = max(contours, key=cv2.contourArea) | |
| if cv2.contourArea(largest) < (h * w) * 0.15: | |
| return None | |
| for source in [largest, cv2.convexHull(largest)]: | |
| epsilon = 0.02 * cv2.arcLength(source, True) | |
| approx = cv2.approxPolyDP(source, epsilon, True) | |
| if len(approx) == 4: | |
| return _apply_perspective(img, approx.reshape(4, 2)) | |
| return None | |
| def _method3_fallback(img): | |
| h, w = img.shape[:2] | |
| scale = 800 / w | |
| return cv2.resize(img, (800, int(h * scale))) | |
| def rectify_scoresheet(img: np.ndarray) -> tuple[np.ndarray, str]: | |
| """Rectifica la perspectiva de una planilla. Devuelve (imagen_rectificada, metodo_usado).""" | |
| result = _method1_paper_edges(img) | |
| if result is not None: | |
| return result, "metodo1_bordes_papel" | |
| result = _method2_grid_lines(img) | |
| if result is not None: | |
| return result, "metodo2_grilla" | |
| return _method3_fallback(img), "metodo3_fallback" | |