def run(self):
print 'BlobFetcher started'
while True:
db_inds = self._get_next_minibatch_inds()
minibatch_db = [self._roidb[i] for i in db_inds]
blobs = get_minibatch(minibatch_db, self._num_classes)
self._queue.put(blobs)
def run(self):
print 'BlobFetcher started'
while True:
db_inds = self._get_next_minibatch_inds()
minibatch_db = [self._roidb[i] for i in db_inds]
blobs = get_minibatch(minibatch_db)
self._queue.put(blobs)
def predict():
#imdb, roidb = combined_roidb('voc_2007_trainval')
imdb, roidb = combined_roidb('voc_2007_test', is_training=False)
print imdb.classes
#from data_layer.layer import RoIDataLayer
#data_layer = RoIDataLayer(roidb, imdb.num_classes)
#for _ in range(10):
# blobs = data_layer.forward()
# print blobs
sv = Predict(imdb.num_classes)
ids=0
for roi in roidb:
print "====================================================================="
print roi
blobs=get_minibatch([roi], imdb.num_classes)
image = blobs['data']
gt_boxes = blobs['gt_boxes']
im_info = blobs['im_info']
clss_score, clss_bbox = sv.predict(image, im_info)
for i,(score, bbox) in enumerate(zip(clss_score, clss_bbox)):
if len(score)>0:
print ("class:",i+1, imdb.classes[i+1])
print "score:",score
print "bbox:",bbox
ids+=1
if ids > 10:
break
def __getitem__(self, item):
minibatch_db = [self.roidb[item]]
blobs = get_minibatch(minibatch_db, self.num_classes)
data = blobs['data']
im_info = blobs['im_info']
if self.training:
gt_boxes = blobs['gt_boxes']
else:
gt_boxes = torch.FloatTensor([0, 0, 0, 0, 0])
return data[0].transpose(2, 0, 1), im_info.reshape([3]), gt_boxes
def _get_next_minibatch(self):
"""Return the blobs to be used for the next minibatch.
If cfg.TRAIN.USE_PREFETCH is True, then blobs will be computed in a
separate process and made available through self._blob_queue.
"""
if cfg.TRAIN.USE_PREFETCH:
return self._blob_queue.get()
else:
db_inds = self._get_next_minibatch_inds()
minibatch_db = [self._roidb[i] for i in db_inds]
return get_minibatch(minibatch_db)
def _get_next_minibatch(self):
"""Return the blobs to be used for the next minibatch.
If cfg.TRAIN.USE_PREFETCH is True, then blobs will be computed in a
separate process and made available through self._blob_queue.
"""
if cfg.TRAIN.USE_PREFETCH:
return self._blob_queue.get()
else:
db_inds = self._get_next_minibatch_inds()
minibatch_db = [self._roidb[i] for i in db_inds]
return get_minibatch(minibatch_db, self._num_classes)
def run(self):
print 'BlobFetcher started'
while True:
minibatch_inds = self._get_next_minibatch_inds()
minibatch_img_paths = \
[self._db.get_img_path(self._db.train_ind[
i if i < len(self._db.train_ind) else i - len(self._db.train_ind)])
for i in minibatch_inds]
minibatch_labels = \
[self._db.labels[self._db.train_ind[
i if i < len(self._db.train_ind) else i - len(self._db.train_ind)]]
for i in minibatch_inds]
minibatch_flip = \
[0 if i < len(self._db.train_ind) else 1
for i in minibatch_inds]
blobs = get_minibatch(minibatch_img_paths, minibatch_labels,
minibatch_flip, self._db.flip_attr_pairs,
self._weight)
self._queue.put(blobs)
def run(self):
print 'BlobFetcher started'
while True:
minibatch_inds, img_ratio = self._get_next_minibatch_inds()
if len(minibatch_inds) < cfg.TRAIN.BATCH_SIZE:
continue
#print "The ratio of this blob is: %d" % img_ratio
#print "Size of minibatch %d" % len(minibatch_inds)
#print minibatch_inds
#print
minibatch_img_paths = \
[self._db.get_img_path(self._db.train_ind[i])
for i in minibatch_inds]
minibatch_labels = \
[self._db.labels[self._db.train_ind[i]]
for i in minibatch_inds]
minibatch_flip = \
np.random.random_integers(0, 1, len(minibatch_inds))
blobs = get_minibatch(minibatch_img_paths, minibatch_labels,
minibatch_flip, self._db.flip_attr_pairs,
self._weight, img_ratio)
self._queue.put(blobs)
def _get_next_minibatch(self):
db_inds = self._get_next_minibatch_inds()
minibatch_db = [self._roidb[i] for i in db_inds]
return get_minibatch(minibatch_db, self._num_classes)
def _get_next_minibatch(self): db_inds = self._get_next_minibatch_inds() # minibatch_db = [self._db[i] for i in db_inds] return get_minibatch(self._db, db_inds, self._labels)
