CNN model evaluation, boxsuppress function how to reconstruct my own mask?

2 次查看（过去 30 天）

显示更早的评论

sana3 sal 2018-6-6

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/404437-cnn-model-evaluation-boxsuppress-function-how-to-reconstruct-my-own-mask

编辑： Image Analyst 2018-6-7

在 MATLAB Online 中打开

Hello, I am working on object detection. I found the code here:

https://github.com/vedaldi/practical-object-category-detection

And I reached the model evaluation step.

I am having a problem with the matrix dimensions.

They use 64*64 image box and a

mask = [-1 0 1 0 ; 0 -1 0 1]

when I use 50*100 image box what should my mask be?

The code is :

function keep = boxsuppress(boxes, scores, threshold)
% BOXSUPPRESS Box non-maxima suprression
%   KEEP = BOXSUPPRESS(BOXES, SCORES, THRESHOLD)
;
scores(any([-1 0 1 0 ; 0 -1 0 1] * boxes < 0)) = -inf ;
keep = false(1, size(boxes,2)) ;
while true
  [score, best] = max(scores) ;
  if score == -inf, break ; end
  keep(best) = true ;
  remove = boxinclusion(boxes(:,best), boxes, 'pascalFormat', true) >= threshold ;
  scores(remove) = -inf ;
  scores(best) = -inf ; % `best` is not in `remove` if threshold > 1
end
end