Add code refinement to this repo

[jakewilliami]

• See Speeding up geting vote and socres of haar feature Simon-Hohberg/Viola-Jones#17.

[jakewilliami]

• See also Simon-Hohberg/Viola-Jones@master...najiji:master (from https://github.com/najiji/Viola-Jones)

[jakewilliami]

• On top of this, need to consider performance: https://docs.julialang.org/en/v1.5/manual/performance-tips/

[jakewilliami]

• Also consider using axes rather than plain referencing (for n dimensional images): https://julialang.org/blog/2017/04/offset-arrays/

That is to say

# Yes:
for x in axes(img_arr, 1)

# No:
for x in 1:arr_rows
for x=1:arr_rows

EDIT: this has been achieved by coords_spatial from Base.

[jakewilliami]

Regarding comment made on 3rd September, these are the commits which I haven't yet assessed (those commits which seem to seriously change the AdaBoost method):

• Simon-Hohberg/Viola-Jones@bd9520a
• Simon-Hohberg/Viola-Jones@0ca1ab6
• Simon-Hohberg/Viola-Jones@ead131b

[jakewilliami]

Assigning @dmipeck to help with this. Not pressing as the algorithm detects faces and non-faces above chance, so the algorithm works, despite the lack of peer review.