Skip to content

Commit

Permalink
- New version of MCG benchmark
Browse files Browse the repository at this point in the history
- This new version implements the measures used in the PAMI submission and in the following paper:
**Jordi Pont-Tuset** and **Luc Van Gool**<br>
Boosting Object Proposals: From Pascal to COCO<br>
*International Conference on Computer Vision (ICCV) 2015*
  • Loading branch information
jponttuset committed Apr 27, 2016
2 parents cfc47fd + 1719f4e commit 591b5a3
Show file tree
Hide file tree
Showing 84 changed files with 918,548 additions and 18,818 deletions.
5 changes: 3 additions & 2 deletions benchmark/README.txt
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,8 @@
% ------------------------------------------------------------------------

FIRST INSTALL
- Change datasets/database_root_dir.m to point to your PASCAL2012 folder (the one with subfolders ImageSets, JPEGImages, etc.) or the SBD or COCO folders.
- Change src/gt_wrappers/db_root_dir.m to point to your PASCAL2012 folder (the one with subfolders ImageSets, JPEGImages, etc.) or the SBD or COCO folders.
Check src/gt_wrappers/README.md for more info on this
- Run install.m from the root dir to add the needed paths and do some checks
- If you need to re-build the library (the script install.m will tell if needed), run build.m

Expand All @@ -24,7 +25,7 @@ USAGE INSTALL
- If you want to avoid this, add the paths permanently

EVALUATION
- See results_segm_proposals.m and results_box_proposals for an example of how to evaluate and show the benchmark results
- See results_segm_proposals.m and results_box_proposals.m for an example of how to evaluate and show the benchmark results

Enjoy!

Expand Down
7 changes: 0 additions & 7 deletions benchmark/build.m
Original file line number Diff line number Diff line change
Expand Up @@ -17,11 +17,7 @@
% This function builds all the MEX files needed.
% Dependencies: Boost C++ libraries (http://www.boost.org)
%
% The code contains C++11 functionality, so you have to use a compiler that
% supports the flag -std=c++11.
% Some help on how to do it in: http://jponttuset.github.io/matlab2014-mex-flags/
% ------------------------------------------------------------------------

function build()
%% Check that root_dir is set properly
if ~exist(root_dir,'dir')
Expand Down Expand Up @@ -55,9 +51,6 @@ function build()
eval(['mex ''' build_file{ii} ''' -outdir ''' fullfile(root_dir, 'lib') '''' include_str])
end

%% Build COCO
eval(['mex src/coco/private/gasonMex.cpp src/coco/private/gason.cpp -Isrc/coco/private -I/usr/local/include -outdir ''' fullfile(root_dir, 'lib') '''' include_str])

%% Clear variables
clear build_file ii include include_str

Expand Down
42 changes: 30 additions & 12 deletions benchmark/datasets/database_root_dir.m → benchmark/check_dbs.m
Original file line number Diff line number Diff line change
Expand Up @@ -13,19 +13,37 @@
% Computer Vision and Pattern Recognition (CVPR) 2014.
% Please consider citing the paper if you use this code.
% ------------------------------------------------------------------------
% Adapt the paths to the place where you have downloaded PASCAL.
% ------------------------------------------------------------------------

function db_root_dir = database_root_dir( database )
if strcmp(database,'pascal2012')
db_root_dir ='/path/to/pascal2012/';
elseif strcmp(database,'COCO')
db_root_dir = '/path/to/COCO/';
elseif strcmp(database,'SBD')
db_root_dir = '/path/to/SBD/';
else
error(['Unknown database: ' database]);
end
%% Get all image IDs from a certain dataset
database = 'COCO';
gt_set = 'train2014';
% database = 'Pascal';
% gt_set = 'Segmentation_train_2012';
% database = 'SBD';
% gt_set = 'train';
sel_id = 2;

% Read image IDs
im_ids = db_ids(database,gt_set);

% Show size and example ID
disp(['There are ' num2str(length(im_ids)) ' images in ' database ' - ' gt_set])
disp([' Example ID ''' im_ids{sel_id} ''' in position ' num2str(sel_id)])

%% Load an image
im = db_im(database,im_ids{sel_id});

% Display
figure; subplot(2,2,1);
imshow(im)

%% Load a ground-truth annotation (first time in COCO takes longer)
gt = db_gt(database,im_ids{sel_id});

for ii=1:min(3,length(gt.masks))
subplot(2,2,1+ii);
imshow(gt.masks{ii})
end



23 changes: 0 additions & 23 deletions benchmark/datasets/database_ids.m

This file was deleted.

101 changes: 0 additions & 101 deletions benchmark/datasets/get_ground_truth.m

This file was deleted.

37 changes: 0 additions & 37 deletions benchmark/datasets/get_image.m

This file was deleted.

Loading

0 comments on commit 591b5a3

Please sign in to comment.