output directory

datasets/VOCdevkit-matlab-wrapper/voc_eval.m

-function res = voc_eval(path, comp_id, test_set, rm_res)
+function res = voc_eval(path, comp_id, test_set, output_dir, rm_res)
 
 VOCopts = get_voc_opts(path);
 VOCopts.testset = test_set;
 
 for i = 1:length(VOCopts.classes)
   cls = VOCopts.classes{i};
-  res(i) = voc_eval_cls(cls, VOCopts, comp_id, rm_res);
+  res(i) = voc_eval_cls(cls, VOCopts, comp_id, output_dir, rm_res);
 end
 
 fprintf('\n~~~~~~~~~~~~~~~~~~~~\n');
@@ -15,7 +15,7 @@ disp(aps);
 disp(mean(aps));
 fprintf('~~~~~~~~~~~~~~~~~~~~\n');
 
-function res = voc_eval_cls(cls, VOCopts, comp_id, rm_res)
+function res = voc_eval_cls(cls, VOCopts, comp_id, output_dir, rm_res)
 
 test_set = VOCopts.testset;
 year = VOCopts.dataset(4:end);
@@ -36,12 +36,12 @@ if do_eval
   [recall, prec, ap] = VOCevaldet(VOCopts, comp_id, cls, true);
   ap_auc = xVOCap(recall, prec);
 
-  % % force plot limits
-  % ylim([0 1]);
-  % xlim([0 1]);
+  % force plot limits
+  ylim([0 1]);
+  xlim([0 1]);
 
-  % print(gcf, '-djpeg', '-r0', ...
-  %     [conf.cache_dir cls '_pr_' imdb.name suffix '.jpg']);
+  print(gcf, '-djpeg', '-r0', ...
+        [output_dir '/' cls '_pr.jpg']);
 end
 fprintf('!!! %s : %.4f %.4f\n', cls, ap, ap_auc);
 
@@ -50,8 +50,8 @@ res.prec = prec;
 res.ap = ap;
 res.ap_auc = ap_auc;
 
-%save([conf.cache_dir cls '_pr_' imdb.name '_py'], ...
-%    'res', 'recall', 'prec', 'ap', 'ap_auc');
+save([output_dir '/' cls '_pr.mat'], ...
+     'res', 'recall', 'prec', 'ap', 'ap_auc');
 
 if rm_res
   delete(res_fn);

datasets/imdb.py

@@ -67,7 +67,7 @@ class imdb(object):
     def default_roidb(self):
         raise NotImplementedError
 
-    def evaluate_detections(self, all_boxes):
+    def evaluate_detections(self, all_boxes, output_dir=None):
         """
         all_boxes is a list of length number-of-classes.
         Each list element is a list of length number-of-images.

datasets/pascal_voc.py

@@ -232,7 +232,7 @@ class pascal_voc(datasets.imdb):
                                        dets[k, 2] + 1, dets[k, 3] + 1))
         return comp_id
 
-    def _do_matlab_eval(self, comp_id):
+    def _do_matlab_eval(self, comp_id, output_dir='output'):
         rm_results = self.config['cleanup']
 
         path = os.path.join(os.path.dirname(__file__),
@@ -240,15 +240,15 @@ class pascal_voc(datasets.imdb):
         cmd = 'cd {} && '.format(path)
         cmd += 'matlab -nodisplay -nodesktop '
         cmd += '-r "dbstop if error; '
-        cmd += 'voc_eval(\'{:s}\', \'{:s}\', \'{:s}\', {:d}); quit;"' \
+        cmd += 'voc_eval(\'{:s}\',\'{:s}\',\'{:s}\',\'{:s}\',{:d}); quit;"' \
                .format(self._devkit_path, comp_id,
-                       self._image_set, int(rm_results))
+                       self._image_set, output_dir, int(rm_results))
         print('Running:\n{}'.format(cmd))
         status = subprocess.call(cmd, shell=True)
 
-    def evaluate_detections(self, all_boxes):
+    def evaluate_detections(self, all_boxes, output_dir):
         comp_id = self._write_voc_results_file(all_boxes)
-        self._do_matlab_eval(comp_id)
+        self._do_matlab_eval(comp_id, output_dir)
 
 if __name__ == '__main__':
     d = datasets.pascal_voc('trainval', '2007')

experiments/test_net.py

@@ -4,6 +4,7 @@ import fast_rcnn_test
 import caffe
 import argparse
 import sys
+import os
 
 def parse_args():
     """
@@ -33,6 +34,7 @@ if __name__ == '__main__':
     caffe.set_mode_gpu()
     caffe.set_device(args.gpu_id)
     net = caffe.Net(args.prototxt, args.caffemodel, caffe.TEST)
+    net.name = os.path.splitext(os.path.basename(args.caffemodel))[0]
 
     import datasets.pascal_voc
     imdb = datasets.pascal_voc('test', '2007')

fast_rcnn_test.py

@@ -10,6 +10,7 @@ import utils.cython_nms
 import cPickle
 import heapq
 import utils.blob
+import os
 
 def _get_image_blob(im):
     im_pyra = []
@@ -204,6 +205,13 @@ def test_net(net, imdb):
     all_boxes = [[[] for _ in xrange(num_images)]
                  for _ in xrange(imdb.num_classes)]
 
+    # Output directory will be something like:
+    #   output/vgg16_fast_rcnn_iter_40000/voc_2007_test/
+    output_dir = os.path.join(os.path.dirname(__file__), 'output',
+                              net.name, imdb.name)
+    if not os.path.exists(output_dir):
+        os.makedirs(output_dir)
+
     # timers
     _t = {'im_detect' : Timer(), 'misc' : Timer()}
 
@@ -251,12 +259,12 @@ def test_net(net, imdb):
             inds = np.where(all_boxes[j][i][:, -1] > thresh[j])[0]
             all_boxes[j][i] = all_boxes[j][i][inds, :]
 
-    # TODO(rbg): need to have an output directory to save results in
-    with open('dets.pkl', 'wb') as f:
+    det_file = os.path.join(output_dir, 'detections.pkl')
+    with open(det_file, 'wb') as f:
         cPickle.dump(all_boxes, f, cPickle.HIGHEST_PROTOCOL)
 
     print 'Applying NMS to all detections'
     nms_dets = _apply_nms(all_boxes, conf.TEST_NMS)
 
     print 'Evaluating detections'
-    imdb.evaluate_detections(nms_dets)
+    imdb.evaluate_detections(nms_dets, output_dir)