Python OpenCV RTrees does not load properly


I have a trained RTrees model using OpenCV version 3.3.1 which I have saved and need to later load for prediction.

I am attempting to use OpenCV’s RTrees to create a random forest classifier. However, I am not able to successfully load a model saved using the save function.
An example of the code I used is as follows:

import numpy as np
import cv2

def train(samples, class_labels, save_file=None):
    samples      : np.ndarray of type np.float32
    class_labels :  np.ndarray of type int
    save_file    : str of the absolute path to the save file
    model =

    # Set paremeters
    term_type, n_trees, epsilon = cv2.TERM_CRITERIA_MAX_ITER, 128, 1
    model.setTermCriteria((term_type, n_trees, epsilon))

    train_data =,

    if save_file:

def test(samples, load_file):
        samples   : np.ndarray of type np.float32
        load_file : str of the absolute path to the load file
        model =

        _ret, responses = model.predict(samples)
        return responses.ravel()

if __name__ == '__main__':
    model_file = '/home/user/bin/tmp/m.xml'

    x = np.random.randint(0, 5, (1000, 15))
    y = np.random.randint(0, 4, 1000)
    train(x, y, model_file)

    x_test = np.random.randint(0, 5, (100, 15))
    test(x_test, model_file)

The error is as follows:

Traceback (most recent call last):
  File "", line 45, in <module>
    test(x_test, model_file)
  File "", line 34, in test
    _ret, responses = model.predict(samples)
cv2.error: OpenCV(3.4.2) /io/opencv/modules/ml/src/tree.cpp:1498: error: (-215:Assertion failed) !roots.empty() in function 'predict'
Asked By: patyx



The issue is that the load function is a static function which returns the loaded model.
OpenCV’s documentation and examples do not specifically indicate this.

model =
model = model.load(file_name)

# Or:
model =

See this on the OpenCV site.

Answered By: patyx