How to sum number of occurrences within an np array efficiently? Area of mask RCNN

Question

I'm working with the library MaskRCNN and i want to get the area of each mask. All the masks are an array (W x H) with values False or True. All the trues represent a mask and all the falses a background.

#mask 0
r['masks'][:,:,0] 
#output-->
   array([[False, False, False, ..., False, False, False],
   [False, False, False, ..., False, False, False],
   [False, False, False, ..., False, False, False],
   ...,
   [False, False, False, ..., False, False, False],
   [False, False, False, ..., False, False, False],
   [False, False, False, ..., False, False, False]])

How can i iterate over all and sum all the true occurrences, both in columns and rows?

This i suppose give me the area in pixels of the mask, which i can interpolate real area.

Give us [`mcve`](https://stackoverflow.com/help/minimal-reproducible-example)? — Divakar, Sep 04 '19 at 09:44
This guy had a similar problem: https://stackoverflow.com/questions/28663856/how-to-count-the-occurrence-of-certain-item-in-an-ndarray-in-python — Claudio Busatto, Sep 04 '19 at 09:46

score 1 · Accepted Answer · answered Sep 04 '19 at 09:51

Based on this previous question, you need first to flatten the numpy array and therefore apply the same idea, i.e.,

flat_r = numpy.flatten(r['masks'][:,:,0])

unique, counts = numpy.unique(flat_r, return_counts=True)

print(dict(zip(unique, counts))) 
# {False: 7, True: 4}

score 0 · Answer 2 · answered Sep 04 '19 at 21:38

0

Claudio answer was good, but i found a simple way too.

true_occurrences= np.sum(r['masks'][:,:,0])

answered Sep 04 '19 at 21:38

hopieman

399
7
22

How to sum number of occurrences within an np array efficiently? Area of mask RCNN

2 Answers2