paperdoll icon indicating copy to clipboard operation
paperdoll copied to clipboard

Request help "Error:feature_calculator"

Open parez93 opened this issue 8 years ago • 3 comments

I need some help for running paperdoll (pre-trained parser for a new image).

When I run: load data/paperdoll_pipeline.mat config; input_image = imread('./my_image.jpg'); input_sample = struct('image', imencode(input_image, 'jpg')); result = feature_calculator.apply(config, input_sample); %error

The result comes:

[06-Nov-2016 06:29:48] Error:feature_calculator: 1 / 1
Attempted to access pose_struct(1,:); index out of bounds because size(pose_struct)=[0,0].

Error in normal_pose_calculator.apply (line 40)
      pose_struct = pose.PARSE_from_UCI(pose.box2point(pose_struct(1,:)));

Error in feature_calculator.apply (line 25)
        sample = calculators{j}(config{j}, sample, varargin{:}, 'Encode', false);


result =

     []

Is there any solution to fix it ?

While platforms are:

  • ubuntu 12.04 64 bit
  • Matlab R2012a
  • opencv 2.4.11
  • no error when i run makefile

parez93 avatar Nov 06 '16 14:11 parez93

@parez93 You'll get this when pose estimator fail to find a person. Try lowering the pose threshold as indicated in the README.

kyamagu avatar Nov 06 '16 14:11 kyamagu

Thank you very much! I've another question: when i run show_parsing(result.image, result.final_labeling, result.refined_labels); I get this error:

[`Error using horzcat
Dimensions of matrices being concatenated are not consistent.

Error in show_parsing>reorder_labels (line 23)
  ordered_labels = [reserved_labels, setdiff(labels, reserved_labels)];

Error in show_parsing (line 12)
  [labeling, labels] = reorder_labels(uint8(labeling), labels);

Maybe it is a my mistake, but if i change in "show_parsing.m" function, all go right: ordered_labels = [reserved_labels, setdiff(labels, reserved_labels)]; in ordered_labels = [reserved_labels, setdiff(labels, reserved_labels)']; (i added a single quote at the end).

parez93 avatar Nov 06 '16 15:11 parez93

@parez93 Right, this commit should have fixed this.

kyamagu avatar Nov 06 '16 22:11 kyamagu