Contents

% DEMO: PASCAL VOC "mini" training/testing script
% This function can generate a nice HTML page by calling:
% publish('esvm_demo_train_voc_class_fast.m','html')
%
% Copyright (C) 2011-12 by Tomasz Malisiewicz
% All rights reserved.
%
% This file is part of the Exemplar-SVM library and is made
% available under the terms of the MIT license (see COPYING file).
% Project homepage: https://github.com/quantombone/exemplarsvm
function [models,M] = esvm_demo_train_voc_class_fast(cls, ...
                                                  data_directory, ...
                                                  dataset_directory, ...
                                                  results_directory)
addpath(genpath(pwd));

if ~exist('cls','var')
  fprintf(1,'esvm_demo_train_fast: defaulting to class=car\n');
  cls = 'car';
end

if ~exist('data_directory','var')
  data_directory = '/Users/tomasz/projects/pascal/';
end

if ~exist('dataset_directory','var')
  dataset_directory = 'VOC2007';
end

if ~exist('results_directory','var')
  %results_directory = '';
  results_directory = sprintf(['/nfs/baikal/tmalisie/esvm-%s-' ...
                    '%s-fast/'], ...
                              dataset_directory, cls);
end

%data_directory = '/Users/tomasz/projects/Pascal_VOC/';
%results_directory = '/nfs/baikal/tmalisie/esvm-data/';

%data_directory = '/csail/vision-videolabelme/people/tomasz/VOCdevkit/';
%results_directory = sprintf('/csail/vision-videolabelme/people/tomasz/esvm-%s/',cls);


dataset_params = esvm_get_voc_dataset(dataset_directory,...
                                      data_directory,...
                                      results_directory);


dataset_params.display = 1;
%dataset_params.dump_images = 1;

% Issue warning if lock files are present
lockfiles = check_for_lock_files(results_directory);
if length(lockfiles) > 0
  fprintf(1,'WARNING: %d lockfiles present in current directory\n', ...
          length(lockfiles));
end

%KILL_LOCKS = 1;
%for i = 1:length(lockfiles)
%  unix(sprintf('rmdir %s',lockfiles{i}));
%end
esvm_demo_train_fast: defaulting to class=car

Set exemplar-initialization parameters

params = esvm_get_default_params;
params.model_type = 'exemplar';
params.dataset_params = dataset_params;

%Initialize exemplar stream
stream_params.stream_set_name = 'trainval';
stream_params.stream_max_ex = 1;
stream_params.must_have_seg = 0;
stream_params.must_have_seg_string = '';
stream_params.model_type = 'exemplar'; %must be scene or exemplar;
stream_params.cls = cls;

%Create an exemplar stream (list of exemplars)
e_stream_set = esvm_get_pascal_stream(stream_params, ...
                                      dataset_params);

neg_set = esvm_get_pascal_set(dataset_params, ['train-' cls]);

%Choose a models name to indicate the type of training run we are doing
models_name = ...
    [cls '-' params.init_params.init_type ...
     '.' params.model_type];

initial_models = esvm_initialize_exemplars(e_stream_set, params, models_name);
.Making directory /nfs/baikal/tmalisie/esvm-VOC2007-car-fast///models/car-g.exemplar/
.initialized with HOG_size = [10 12]
Load from a total of 1 files:
.
Loaded models, saving to /nfs/baikal/tmalisie/esvm-VOC2007-car-fast///models//car-g.exemplar.mat

   --- Done initializing 1 exemplars

Perform Exemplar-SVM training

train_params = params;
train_params.detect_max_scale = 0.5;
train_params.train_max_mined_images = 50;
train_params.detect_exemplar_nms_os_threshold = 1.0;
train_params.detect_max_windows_per_exemplar = 100;

Train the exemplars and get updated models name

[models,models_name] = esvm_train_exemplars(initial_models, ...
                                            neg_set, train_params);


val_params = params;
val_params.detect_exemplar_nms_os_threshold = 0.5;
val_params.gt_function = @esvm_load_gt_function;

val_set_name = ['trainval+' cls];

val_set = esvm_get_pascal_set(dataset_params, val_set_name);
val_set = val_set(1:40);
Randomizing mining queue
Found 0200 windows, image:01226 (#seen=00001/02099, max = 25.719)
Found 0200 windows, image:01157 (#seen=00002/02099, max = 20.872)
Found 0200 windows, image:00416 (#seen=00003/02099, max = 19.339)
Found 0200 windows, image:00105 (#seen=00004/02099, max = 20.110)
Found 0200 windows, image:00388 (#seen=00005/02099, max = 19.025)
Stopping mining because we have 1000 windows from 5 new violators
# Violating images: 5, #Non-violating images: 0
 -----
Starting SVM: dim=3720... #pos=1, #neg=1000  --- Max positive is 1.000
SVM iteration took 1.236 sec,  kept 84 negatives
Found 0200 windows, image:02091 (#seen=00006/02099, max = -0.785)
Found 0000 windows, image:00884 (#seen=00007/02099)
Found 0130 windows, image:00270 (#seen=00008/02099, max = -0.840)
Found 0183 windows, image:01721 (#seen=00009/02099, max = -0.802)
Found 0088 windows, image:00943 (#seen=00010/02099, max = -0.894)
Found 0116 windows, image:00148 (#seen=00011/02099, max = -0.799)
Found 0030 windows, image:00326 (#seen=00012/02099, max = -0.929)
Found 0056 windows, image:00763 (#seen=00013/02099, max = -0.847)
Found 0129 windows, image:00513 (#seen=00014/02099, max = -0.885)
Found 0200 windows, image:01258 (#seen=00015/02099, max = -0.531)
Stopping mining because we have 1132 windows from 9 new violators
# Violating images: 9, #Non-violating images: 1
 -----
Starting SVM: dim=3720... #pos=1, #neg=1216  --- Max positive is 1.000
SVM iteration took 1.408 sec,  kept 75 negatives
Found 0003 windows, image:00374 (#seen=00016/02099, max = -0.985)
Found 0064 windows, image:00619 (#seen=00017/02099, max = -0.849)
Found 0108 windows, image:00824 (#seen=00018/02099, max = -0.786)
Found 0039 windows, image:00774 (#seen=00019/02099, max = -0.946)
Found 0047 windows, image:01514 (#seen=00020/02099, max = -0.868)
Found 0018 windows, image:01117 (#seen=00021/02099, max = -0.962)
Found 0006 windows, image:01647 (#seen=00022/02099, max = -0.935)
Found 0011 windows, image:01594 (#seen=00023/02099, max = -0.911)
Found 0000 windows, image:01344 (#seen=00024/02099)
Found 0010 windows, image:00982 (#seen=00025/02099, max = -0.954)
Found 0108 windows, image:00575 (#seen=00026/02099, max = -0.850)
Found 0155 windows, image:01266 (#seen=00027/02099, max = -0.807)
Found 0000 windows, image:01062 (#seen=00028/02099)
Found 0025 windows, image:00213 (#seen=00029/02099, max = -0.935)
Found 0023 windows, image:01489 (#seen=00030/02099, max = -0.915)
Found 0000 windows, image:01419 (#seen=00031/02099)
Found 0197 windows, image:02075 (#seen=00032/02099, max = -0.784)
Found 0009 windows, image:01121 (#seen=00033/02099, max = -0.878)
Found 0059 windows, image:01372 (#seen=00034/02099, max = -0.900)
Found 0010 windows, image:01791 (#seen=00035/02099, max = -0.953)
Found 0000 windows, image:01525 (#seen=00036/02099)
Found 0200 windows, image:01234 (#seen=00037/02099, max = -0.801)
Stopping mining because we have 1092 windows from 18 new violators
# Violating images: 18, #Non-violating images: 4
 -----
Starting SVM: dim=3720... #pos=1, #neg=1167  --- Max positive is 1.000
SVM iteration took 1.502 sec,  kept 105 negatives
Found 0019 windows, image:00343 (#seen=00038/02099, max = -0.881)
Found 0000 windows, image:00872 (#seen=00039/02099)
Found 0000 windows, image:00789 (#seen=00040/02099)
Found 0000 windows, image:02055 (#seen=00041/02099)
Found 0039 windows, image:01094 (#seen=00042/02099, max = -0.910)
Found 0019 windows, image:00507 (#seen=00043/02099, max = -0.919)
Found 0000 windows, image:00909 (#seen=00044/02099)
Found 0002 windows, image:01635 (#seen=00045/02099, max = -0.997)
Found 0001 windows, image:01485 (#seen=00046/02099, max = -0.999)
Found 0000 windows, image:01456 (#seen=00047/02099)
Found 0100 windows, image:01973 (#seen=00048/02099, max = -0.778)
Found 0000 windows, image:02044 (#seen=00049/02099)
Found 0004 windows, image:00933 (#seen=00050/02099, max = -0.968)
Stopping mining because we have 184 windows from 7 new violators
# Violating images: 7, #Non-violating images: 6
 -----
Starting SVM: dim=3720... #pos=1, #neg=289  --- Max positive is 1.000
SVM iteration took 0.998 sec,  kept 105 negatives
Found 0000 windows, image:01082 (#seen=00051/02099)
Stopping mining because we have 0 windows from 0 new violators
# Violating images: 0, #Non-violating images: 1
 -----
Starting SVM: dim=3720... #pos=1, #neg=105  --- Max positive is 1.000
SVM iteration took 0.946 sec,  kept 105 negatives
 ### End of training... 
Load from a total of 1 files:
.
Loaded models, saving to /nfs/baikal/tmalisie/esvm-VOC2007-car-fast///models//car-g.exemplar-svm.mat
Saving stripped to /nfs/baikal/tmalisie/esvm-VOC2007-car-fast///models//car-g.exemplar-svm-stripped.mat

Apply trained exemplars on validation set

val_grid = esvm_detect_imageset(val_set, models, val_params, val_set_name);
Making directory /nfs/baikal/tmalisie/esvm-VOC2007-car-fast///detections/trainval+car-car-g.exemplar-svm/
 --image 00001/00040: 1 exemplars took 0.740sec, #windows=00003, max=-0.854 
 --image 00002/00040: 1 exemplars took 0.830sec, #windows=00002, max=-0.851 
 --image 00003/00040: 1 exemplars took 0.445sec, #windows=00003, max=-0.961 
 --image 00004/00040: 1 exemplars took 0.952sec, #windows=00002, max=-0.809 
 --image 00005/00040: 1 exemplars took 0.709sec, #windows=00002, max=-0.811 
 --image 00006/00040: 1 exemplars took 0.887sec, #windows=00004, max=-0.878 
 --image 00007/00040: 1 exemplars took 0.849sec, #windows=00002, max=-0.967 
 --image 00008/00040: 1 exemplars took 0.813sec, #windows=00006, max=-0.878 
 --image 00009/00040: 1 exemplars took 0.622sec, #windows=00003, max=1.000 
 --image 00010/00040: 1 exemplars took 0.599sec, #windows=00002, max=-0.955 
 --image 00011/00040: 1 exemplars took 1.024sec, #windows=00006, max=-0.795 
 --image 00012/00040: 1 exemplars took 0.856sec, #windows=00005, max=-0.823 
 --image 00013/00040: 1 exemplars took 0.599sec, #windows=00005, max=-0.851 
 --image 00014/00040: 1 exemplars took 0.680sec, #windows=00004, max=-0.835 
 --image 00015/00040: 1 exemplars took 0.950sec, #windows=00004, max=-0.850 
 --image 00016/00040: 1 exemplars took 0.857sec, #windows=00002, max=-0.960 
 --image 00017/00040: 1 exemplars took 0.602sec, #windows=00005, max=-0.875 
 --image 00018/00040: 1 exemplars took 0.663sec, #windows=00003, max=-0.851 
 --image 00019/00040: 1 exemplars took 1.126sec, #windows=00004, max=-0.783 
 --image 00020/00040: 1 exemplars took 0.699sec, #windows=00004, max=-0.945 
 --image 00021/00040: 1 exemplars took 0.713sec, #windows=00003, max=-0.902 
 --image 00022/00040: 1 exemplars took 0.800sec, #windows=00002, max=-0.780 
 --image 00023/00040: 1 exemplars took 0.659sec, #windows=00004, max=-0.722 
 --image 00024/00040: 1 exemplars took 0.672sec, #windows=00004, max=-0.875 
 --image 00025/00040: 1 exemplars took 1.040sec, #windows=00002, max=-0.733 
 --image 00026/00040: 1 exemplars took 0.761sec, #windows=00002, max=-0.827 
 --image 00027/00040: 1 exemplars took 0.536sec, #windows=00000, max= 
 --image 00028/00040: 1 exemplars took 0.845sec, #windows=00006, max=-0.947 
 --image 00029/00040: 1 exemplars took 0.897sec, #windows=00003, max=-0.616 
 --image 00030/00040: 1 exemplars took 0.621sec, #windows=00003, max=-0.878 
 --image 00031/00040: 1 exemplars took 0.730sec, #windows=00004, max=-0.949 
 --image 00032/00040: 1 exemplars took 0.989sec, #windows=00004, max=-0.909 
 --image 00033/00040: 1 exemplars took 0.725sec, #windows=00004, max=-0.924 
 --image 00034/00040: 1 exemplars took 0.786sec, #windows=00002, max=-0.825 
 --image 00035/00040: 1 exemplars took 0.885sec, #windows=00003, max=-0.787 
 --image 00036/00040: 1 exemplars took 0.616sec, #windows=00004, max=-0.809 
 --image 00037/00040: 1 exemplars took 0.690sec, #windows=00003, max=-0.625 
 --image 00038/00040: 1 exemplars took 1.103sec, #windows=00003, max=-0.742 
 --image 00039/00040: 1 exemplars took 0.760sec, #windows=00001, max=-0.949 
 --image 00040/00040: 1 exemplars took 0.595sec, #windows=00005, max=-0.770 

Perform Platt calibration and M-matrix estimation

M = esvm_perform_calibration(val_grid, val_set, models, ...
                             val_params);
.
Loaded calibration parameters "betas", saving to /nfs/baikal/tmalisie/esvm-VOC2007-car-fast///models/car-g.exemplar-svm-betas.mat
 -Computing Box Features:......................................took 0.049sec
 -Learning M by counting: took 0.004sec
 -Applying M to 38 images: took 0.005sec
Computed M, saving to /nfs/baikal/tmalisie/esvm-VOC2007-car-fast///models/car-g.exemplar-svm-M.mat

Define test-set

test_params = params;
test_params.detect_exemplar_nms_os_threshold = 0.5;
test_set_name = ['test+' cls];
test_set = esvm_get_pascal_set(dataset_params, test_set_name);
test_set = test_set(1:100);

Apply on test set

test_grid = esvm_detect_imageset(test_set, models, test_params, test_set_name);
Making directory /nfs/baikal/tmalisie/esvm-VOC2007-car-fast///detections/test+car-car-g.exemplar-svm/
 --image 00001/00100: 1 exemplars took 0.866sec, #windows=00004, max=-0.865 
 --image 00002/00100: 1 exemplars took 1.070sec, #windows=00005, max=-0.940 
 --image 00003/00100: 1 exemplars took 0.588sec, #windows=00002, max=-0.772 
 --image 00004/00100: 1 exemplars took 0.676sec, #windows=00003, max=-0.829 
 --image 00005/00100: 1 exemplars took 0.866sec, #windows=00002, max=-0.719 
 --image 00006/00100: 1 exemplars took 0.895sec, #windows=00005, max=-0.837 
 --image 00007/00100: 1 exemplars took 0.608sec, #windows=00003, max=-0.906 
 --image 00008/00100: 1 exemplars took 0.568sec, #windows=00001, max=-0.780 
 --image 00009/00100: 1 exemplars took 0.513sec, #windows=00003, max=-0.815 
 --image 00010/00100: 1 exemplars took 0.574sec, #windows=00004, max=-0.821 
 --image 00011/00100: 1 exemplars took 0.609sec, #windows=00005, max=-0.817 
 --image 00012/00100: 1 exemplars took 0.531sec, #windows=00004, max=-0.967 
 --image 00013/00100: 1 exemplars took 0.568sec, #windows=00002, max=-0.907 
 --image 00014/00100: 1 exemplars took 0.510sec, #windows=00004, max=-0.923 
 --image 00015/00100: 1 exemplars took 0.516sec, #windows=00004, max=-0.855 
 --image 00016/00100: 1 exemplars took 0.514sec, #windows=00001, max=-0.822 
 --image 00017/00100: 1 exemplars took 0.423sec, #windows=00001, max=-0.984 
 --image 00018/00100: 1 exemplars took 0.592sec, #windows=00003, max=-0.851 
 --image 00019/00100: 1 exemplars took 0.592sec, #windows=00006, max=-0.879 
 --image 00020/00100: 1 exemplars took 0.573sec, #windows=00007, max=-0.928 
 --image 00021/00100: 1 exemplars took 0.670sec, #windows=00002, max=-0.902 
 --image 00022/00100: 1 exemplars took 0.591sec, #windows=00006, max=-0.944 
 --image 00023/00100: 1 exemplars took 0.548sec, #windows=00003, max=-0.914 
 --image 00024/00100: 1 exemplars took 0.611sec, #windows=00003, max=-0.946 
 --image 00025/00100: 1 exemplars took 0.525sec, #windows=00004, max=-0.803 
 --image 00026/00100: 1 exemplars took 0.531sec, #windows=00005, max=-0.963 
 --image 00027/00100: 1 exemplars took 0.590sec, #windows=00002, max=-0.923 
 --image 00028/00100: 1 exemplars took 0.520sec, #windows=00002, max=-0.887 
 --image 00029/00100: 1 exemplars took 0.469sec, #windows=00003, max=-0.830 
 --image 00030/00100: 1 exemplars took 0.602sec, #windows=00000, max= 
 --image 00031/00100: 1 exemplars took 0.520sec, #windows=00002, max=-0.985 
 --image 00032/00100: 1 exemplars took 0.561sec, #windows=00003, max=-0.815 
 --image 00033/00100: 1 exemplars took 0.594sec, #windows=00004, max=-0.881 
 --image 00034/00100: 1 exemplars took 0.524sec, #windows=00003, max=-0.798 
 --image 00035/00100: 1 exemplars took 0.207sec, #windows=00001, max=-0.924 
 --image 00036/00100: 1 exemplars took 0.596sec, #windows=00003, max=-0.779 
 --image 00037/00100: 1 exemplars took 0.576sec, #windows=00004, max=-0.777 
 --image 00038/00100: 1 exemplars took 0.589sec, #windows=00003, max=-0.798 
 --image 00039/00100: 1 exemplars took 0.585sec, #windows=00006, max=-0.949 
 --image 00040/00100: 1 exemplars took 0.493sec, #windows=00004, max=-0.886 
 --image 00041/00100: 1 exemplars took 0.488sec, #windows=00005, max=-0.955 
 --image 00042/00100: 1 exemplars took 0.613sec, #windows=00003, max=-0.847 
 --image 00043/00100: 1 exemplars took 0.486sec, #windows=00004, max=-0.851 
 --image 00044/00100: 1 exemplars took 0.576sec, #windows=00002, max=-0.688 
 --image 00045/00100: 1 exemplars took 0.545sec, #windows=00005, max=-0.897 
 --image 00046/00100: 1 exemplars took 0.462sec, #windows=00006, max=-0.875 
 --image 00047/00100: 1 exemplars took 0.455sec, #windows=00007, max=-0.915 
 --image 00048/00100: 1 exemplars took 0.462sec, #windows=00004, max=-0.938 
 --image 00049/00100: 1 exemplars took 0.472sec, #windows=00005, max=-0.780 
 --image 00050/00100: 1 exemplars took 0.523sec, #windows=00005, max=-0.713 
 --image 00051/00100: 1 exemplars took 0.363sec, #windows=00001, max=-0.939 
 --image 00052/00100: 1 exemplars took 0.534sec, #windows=00005, max=-0.789 
 --image 00053/00100: 1 exemplars took 0.518sec, #windows=00003, max=-0.979 
 --image 00054/00100: 1 exemplars took 0.532sec, #windows=00004, max=-0.778 
 --image 00055/00100: 1 exemplars took 0.474sec, #windows=00005, max=-0.817 
 --image 00056/00100: 1 exemplars took 0.526sec, #windows=00007, max=-0.857 
 --image 00057/00100: 1 exemplars took 0.464sec, #windows=00001, max=-0.749 
 --image 00058/00100: 1 exemplars took 0.467sec, #windows=00004, max=-0.919 
 --image 00059/00100: 1 exemplars took 0.475sec, #windows=00004, max=-0.809 
 --image 00060/00100: 1 exemplars took 0.536sec, #windows=00003, max=-0.883 
 --image 00061/00100: 1 exemplars took 0.235sec, #windows=00002, max=-0.852 
 --image 00062/00100: 1 exemplars took 0.470sec, #windows=00003, max=-0.847 
 --image 00063/00100: 1 exemplars took 0.549sec, #windows=00003, max=-0.784 
 --image 00064/00100: 1 exemplars took 0.465sec, #windows=00003, max=-0.846 
 --image 00065/00100: 1 exemplars took 0.437sec, #windows=00006, max=-0.709 
 --image 00066/00100: 1 exemplars took 0.472sec, #windows=00002, max=-0.782 
 --image 00067/00100: 1 exemplars took 0.529sec, #windows=00004, max=-0.795 
 --image 00068/00100: 1 exemplars took 0.522sec, #windows=00004, max=-0.841 
 --image 00069/00100: 1 exemplars took 0.473sec, #windows=00004, max=-0.892 
 --image 00070/00100: 1 exemplars took 0.526sec, #windows=00003, max=-0.813 
 --image 00071/00100: 1 exemplars took 0.525sec, #windows=00006, max=-0.904 
 --image 00072/00100: 1 exemplars took 0.545sec, #windows=00004, max=-0.723 
 --image 00073/00100: 1 exemplars took 0.530sec, #windows=00003, max=-0.793 
 --image 00074/00100: 1 exemplars took 0.448sec, #windows=00005, max=-0.889 
 --image 00075/00100: 1 exemplars took 0.478sec, #windows=00005, max=-0.972 
 --image 00076/00100: 1 exemplars took 0.478sec, #windows=00003, max=-0.846 
 --image 00077/00100: 1 exemplars took 0.530sec, #windows=00005, max=-0.907 
 --image 00078/00100: 1 exemplars took 0.523sec, #windows=00003, max=-0.785 
 --image 00079/00100: 1 exemplars took 0.479sec, #windows=00002, max=-0.984 
 --image 00080/00100: 1 exemplars took 0.461sec, #windows=00003, max=-0.767 
 --image 00081/00100: 1 exemplars took 0.554sec, #windows=00003, max=-0.933 
 --image 00082/00100: 1 exemplars took 0.452sec, #windows=00002, max=-0.816 
 --image 00083/00100: 1 exemplars took 0.534sec, #windows=00003, max=-0.774 
 --image 00084/00100: 1 exemplars took 0.530sec, #windows=00003, max=-0.845 
 --image 00085/00100: 1 exemplars took 0.462sec, #windows=00004, max=-0.813 
 --image 00086/00100: 1 exemplars took 0.484sec, #windows=00006, max=-0.839 
 --image 00087/00100: 1 exemplars took 0.533sec, #windows=00004, max=-0.633 
 --image 00088/00100: 1 exemplars took 0.536sec, #windows=00006, max=-0.958 
 --image 00089/00100: 1 exemplars took 0.538sec, #windows=00004, max=-0.867 
 --image 00090/00100: 1 exemplars took 0.520sec, #windows=00003, max=-0.874 
 --image 00091/00100: 1 exemplars took 0.472sec, #windows=00001, max=-0.867 
 --image 00092/00100: 1 exemplars took 0.514sec, #windows=00005, max=-0.863 
 --image 00093/00100: 1 exemplars took 0.542sec, #windows=00006, max=-0.914 
 --image 00094/00100: 1 exemplars took 0.526sec, #windows=00003, max=-0.773 
 --image 00095/00100: 1 exemplars took 0.516sec, #windows=00002, max=-0.698 
 --image 00096/00100: 1 exemplars took 0.516sec, #windows=00001, max=-0.931 
 --image 00097/00100: 1 exemplars took 0.522sec, #windows=00004, max=-0.916 
 --image 00098/00100: 1 exemplars took 0.582sec, #windows=00003, max=-0.841 
 --image 00099/00100: 1 exemplars took 0.462sec, #windows=00003, max=-0.819 
 --image 00100/00100: 1 exemplars took 0.532sec, #windows=00004, max=-0.852 

Apply calibration matrix to test-set results

test_struct = esvm_pool_exemplar_dets(test_grid, models, M, test_params);
Applying M-matrix to 100 images:....................................................................................................took 0.028sec
Applying NMS (OS thresh=0.300)

Show top detections

maxk = 20;
allbbs = esvm_show_top_dets(test_struct, test_grid, test_set, models, ...
                       params,  maxk, test_set_name);
Showing detection # 1, score=1.799
Showing detection # 2, score=1.480
Showing detection # 3, score=1.427
Showing detection # 4, score=1.405
Showing detection # 5, score=1.382
Showing detection # 6, score=1.377
Showing detection # 7, score=1.360
Showing detection # 8, score=1.117
Showing detection # 9, score=1.110
Showing detection # 10, score=1.086
Showing detection # 11, score=1.035
Showing detection # 12, score=1.022
Showing detection # 13, score=1.021
Showing detection # 14, score=1.007
Showing detection # 15, score=1.005
Showing detection # 16, score=0.992
Showing detection # 17, score=0.989
Showing detection # 18, score=0.980
Showing detection # 19, score=0.968
Showing detection # 20, score=0.940

Perform the exemplar evaluation

[results] = esvm_evaluate_pascal_voc(test_struct, test_grid, params, ...
                                     test_set_name, cls, models_name);
Writing File /nfs/baikal/tmalisie/esvm-VOC2007-car-fast///results//car-g.exemplar-svm-calibrated-M/comp3_det_test+car.txt
car: pr: load: 129/4952
car: pr: load: 273/4952
car: pr: load: 420/4952
car: pr: load: 574/4952
car: pr: load: 707/4952
car: pr: load: 842/4952
car: pr: load: 981/4952
car: pr: load: 1129/4952
car: pr: load: 1272/4952
car: pr: load: 1397/4952
car: pr: load: 1537/4952
car: pr: load: 1668/4952
car: pr: load: 1803/4952
car: pr: load: 1935/4952
car: pr: load: 2071/4952
car: pr: load: 2208/4952
car: pr: load: 2347/4952
car: pr: load: 2489/4952
car: pr: load: 2631/4952
car: pr: load: 2769/4952
car: pr: load: 2886/4952
car: pr: load: 3022/4952
car: pr: load: 3153/4952
car: pr: load: 3268/4952
car: pr: load: 3407/4952
car: pr: load: 3539/4952
car: pr: load: 3676/4952
car: pr: load: 3803/4952
car: pr: load: 3932/4952
car: pr: load: 4069/4952
car: pr: load: 4204/4952
car: pr: load: 4343/4952
car: pr: load: 4473/4952
car: pr: load: 4610/4952
car: pr: load: 4741/4952
car: pr: load: 4870/4952
car: pr: evaluating detections
Time for computing AP: 0.012sec
Just Wrote /nfs/baikal/tmalisie/esvm-VOC2007-car-fast///www/car-g.exemplar-svm-calibrated-M-on-test+car.pdf

ans = 

    [1x1 struct]