Error performing convexHull on MSER detected regions

I have found an example of MSER in Python OpenCV. When I try to run it I get an error when it tries to reshape a list/numpy.array. The error is:

AttributeError: 'list' object has no attribute 'reshape'

How can I fix this issue? In the below simple code I have commented where the error occurs:

import cv2
import numpy as np

img = cv2.imread('../images/Capture2.JPG', 0);
vis = img.copy()
mser = cv2.MSER_create()
regions = mser.detectRegions(img)

hulls = []
for p in regions:
    # Error on below line: 'AttributeError: 'list' object has no attribute 'reshape''
    hulls.append( cv2.convexHull(p.reshape(-1, 1, 2)) )

    # Note a numpy array isn't working either: error: 'ValueError: cannot reshape array of size 2605 into shape (1,2)'
    p = np.array(p)
    hulls.append( cv2.convexHull(p.reshape(-1, 1, 2)) ) 

cv2.polylines(vis, hulls, 1, (0, 255, 0))
cv2.imshow('img', vis)
cv2.waitKey(0)
cv2.destroyAllWindows()

Solution

I think the mser.detectRegions(img) has evolved a bit over various release versions. In my case I have OpenCV version:

import cv2
print cv2.__version__
>>> 3.3.0

And the mser.detectRegions(img) is returning a tupe with two values instead of a single return value. You can fix this by ignoring the second value of tuple:

regions, _ = mser.detectRegions(img)

I have raised the point of version because a lot of examples available online use regions = mser.detectRegions(img). Which may lead to confusion.

As of now I am not sure of the version which caused this ambiguity, so I am suggesting a quick fix to your problem.