Permalink
Cannot retrieve contributors at this time
Fetching contributors…
Cannot retrieve contributors at this time
| import numpy as np | |
| import random | |
| class BadFitShape(Exception): | |
| pass | |
| def fit_line(xs, ys): | |
| a, b = np.polyfit(np.array(xs), np.array(ys), 1) | |
| if abs(a) > 0.5: raise BadFitShape() | |
| return a, b | |
| def fit_parabola(xs, ys): | |
| a, b, c = np.polyfit(xs, ys, 2) | |
| if a < 0: raise BadFitShape() | |
| return a, b, c | |
| def ransac_parabola(points_l, points_r, ransac_iters_max=5, refine_iters_max=2, max_err=2, debug=False): | |
| if len(points_l) < 3 or len(points_r) < 3: return None | |
| points = np.concatenate([points_l, points_r]) | |
| pts_x, pts_y = np.split(points, 2, axis=1) | |
| pts_x, pts_y = np.squeeze(pts_x), np.squeeze(pts_y) | |
| best_parabola = None | |
| best_support = float('-inf') | |
| # Not enough points to start process | |
| if len(points_r) < 3 or len(points_l) < 3: return None | |
| # Perform N RANSAC iterations | |
| for _ in range(ransac_iters_max): | |
| try: | |
| sample = random.sample(points_r, 3) + (random.sample(points_l, 3)) | |
| sample_xs, sample_ys = [x for (x, _) in sample], [y for (_, y) in sample] | |
| a, b, c = fit_parabola(sample_xs, sample_ys) | |
| # Iteratively refine inliers further | |
| for _ in range(refine_iters_max): | |
| pts_parabola_y = a * pts_x ** 2 + b * pts_x + c | |
| inlier_inds = np.squeeze([np.abs(pts_y - pts_parabola_y) < max_err]) | |
| inliers = points[inlier_inds] | |
| if inliers.size < 5: break | |
| a, b, c = fit_parabola(pts_x[inlier_inds], pts_y[inlier_inds]) | |
| support = len(inliers) | |
| if support > best_support: | |
| best_parabola = (a, b, c) | |
| best_support = support | |
| except BadFitShape: continue | |
| # Early termination for > 90% inliers | |
| if len(inliers) / len(points) > 0.90: break | |
| return best_parabola | |
| __lower_eyelid_inliers_min = 30 | |
| def ransac_line(points, ransac_iters_max=5, refine_iters_max=2, max_err=2, debug=False): | |
| if len(points) < __lower_eyelid_inliers_min: return None | |
| pts_x, pts_y = np.split(points, 2, axis=1) | |
| pts_x, pts_y = np.squeeze(pts_x), np.squeeze(pts_y) | |
| best_line = None | |
| best_support = float('-inf') | |
| # Not enough points to start process | |
| if len(points) < __lower_eyelid_inliers_min : return None | |
| # Perform N RANSAC iterations | |
| for _ in range(ransac_iters_max): | |
| try: | |
| sample = random.sample(points, 3) | |
| sample_xs, sample_ys = [x for (x, _) in sample], [y for (_, y) in sample] | |
| a, b = fit_line(sample_xs, sample_ys) | |
| # Iteratively refine inliers further | |
| for _ in range(refine_iters_max): | |
| pts_line_y = a * pts_x + b | |
| inlier_inds = np.squeeze([np.abs(pts_y - pts_line_y) < max_err]) | |
| inliers = points[inlier_inds] | |
| if len(inliers) < 3: break | |
| a, b = fit_line(pts_x[inlier_inds], pts_y[inlier_inds]) | |
| support = len(inliers) | |
| if len(inliers) > __lower_eyelid_inliers_min and support > best_support: | |
| best_line = (a, b) | |
| best_support = support | |
| except BadFitShape: continue | |
| # Early termination for > 90% inliers | |
| if len(inliers) / len(points) > 0.90: break | |
| return best_line |