Search code examples
pythonscikit-image

Skimage RANSACRegressor At least 2 input points needed

scikit-image : 0.19.3
code to reproduce error:

from skimage.measure import LineModelND, ransac
import numpy as np

pts = np.array([[579,298],
 [604,276],
 [354,241],
 [156,240],
 [124,157],
 [625,105]])

model, inliers = ransac(pts, LineModelND, min_samples=4,residual_threshold=2, max_trials=1000)

error: At least 2 input points needed. at file skimage/measure/fit.py line 92
I haven't faced the same error with other points

Solution

  • The RANSAC procedure could not identify 2 or more inliers in the data. The scikit-image implementation does not present a warning in this situation. You can either increase the residual_threshold or decrease the min_samples to "fix" this behavior.

    Scikit-learn's RANSAC implementation of RANSAC might help you out. Here, the min_samples and residual_threshold inputs are optional.

    from sklearn.linear_model import RANSACRegressor
X = pts[:,0].reshape(-1,1)
y = pts[:,1].reshape(-1,1))
reg = RANSACRegressor(random_state=None,
                      min_samples=None,
                      residual_threshold=None
                     ).fit(X,y)