rcnn_cache_pool5_features.m 3.17 KB
Newer Older
Ross Girshick's avatar
Ross Girshick committed
1
function rcnn_cache_pool5_features(imdb, varargin)
2
3
4
5
6
7
8
9
10
11
12
13
14
15
% rcnn_cache_pool5_features(imdb, varargin)
%   Computes pool5 features and saves them to disk. We compute
%   pool5 features because we can easily compute fc6 and fc7
%   features from them on-the-fly and they tend to compress better
%   than fc6 or fc7 features due to greater sparsity.
%
%   Keys that can be passed in:
%
%   start             Index of the first image in imdb to process
%   end               Index of the last image in imdb to process
%   crop_mode         Crop mode (either 'warp' or 'square')
%   crop_padding      Amount of padding in crop
%   net_file          Path to the Caffe CNN to use
%   cache_name        Path to the precomputed feature cache
Ross Girshick's avatar
Ross Girshick committed
16

Ross Girshick's avatar
Ross Girshick committed
17
18
19
20
21
22
23
24
25
26
% AUTORIGHTS
% ---------------------------------------------------------
% Copyright (c) 2014, Ross Girshick
% 
% This file is part of the R-CNN code and is available 
% under the terms of the Simplified BSD License provided in 
% LICENSE. Please retain this notice and LICENSE if you use 
% this file (or any portion of it) in your project.
% ---------------------------------------------------------

Ross Girshick's avatar
Ross Girshick committed
27
28
29
30
31
32
33
ip = inputParser;
ip.addRequired('imdb', @isstruct);
ip.addOptional('start', 1, @isscalar);
ip.addOptional('end', 0, @isscalar);
ip.addOptional('crop_mode', 'warp', @isstr);
ip.addOptional('crop_padding', 16, @isscalar);
ip.addOptional('net_file', ...
Ross Girshick's avatar
Ross Girshick committed
34
    './data/caffe_nets/finetune_voc_2007_trainval_iter_70k', ...
Ross Girshick's avatar
Ross Girshick committed
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
    @isstr);
ip.addOptional('cache_name', ...
    'v1_finetune_voc_2007_trainval_iter_70000', @isstr);

ip.parse(imdb, varargin{:});
opts = ip.Results;
opts.net_def_file = './model-defs/rcnn_batch_256_output_pool5.prototxt';

image_ids = imdb.image_ids;
if opts.end == 0
  opts.end = length(image_ids);
end

% Where to save feature cache
opts.output_dir = ['./feat_cache/' opts.cache_name '/' imdb.name '/'];
mkdir_if_missing(opts.output_dir);

% Log feature extraction
timestamp = datestr(datevec(now()), 'dd.mmm.yyyy:HH.MM.SS');
diary_file = [opts.output_dir 'rcnn_cache_pool5_features_' timestamp '.txt'];
diary(diary_file);
fprintf('Logging output in %s\n', diary_file);

fprintf('\n\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n');
fprintf('Feature caching options:\n');
disp(opts);
fprintf('~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n\n');

% load the region of interest database
roidb = imdb.roidb_func(imdb);

rcnn_model = rcnn_create_model(opts.net_def_file, opts.net_file);
rcnn_model = rcnn_load_model(rcnn_model);
rcnn_model.detectors.crop_mode = opts.crop_mode;
rcnn_model.detectors.crop_padding = opts.crop_padding;

total_time = 0;
count = 0;
for i = opts.start:opts.end
  fprintf('%s: cache features: %d/%d\n', procid(), i, opts.end);

  save_file = [opts.output_dir image_ids{i} '.mat'];
  if exist(save_file, 'file') ~= 0
    fprintf(' [already exists]\n');
    continue;
  end
  count = count + 1;

  tot_th = tic;

  d = roidb.rois(i);
  im = imread(imdb.image_at(i));

  th = tic;
  d.feat = rcnn_features(im, d.boxes, rcnn_model);
  fprintf(' [features: %.3fs]\n', toc(th));

  th = tic;
  save(save_file, '-struct', 'd');
  fprintf(' [saving:   %.3fs]\n', toc(th));

  total_time = total_time + toc(tot_th);
  fprintf(' [avg time: %.3fs (total: %.3fs)]\n', ...
      total_time/count, total_time);
end