def get_edgeboxes_test(img_id, top_num):
try:
raw_boxes = sio.loadmat(
os.path.join(ENV_PATHS.EDGEBOX_PATH,
str(img_id) + '.mat'))['bbs'][0:top_num, :]
except:
import edge_boxes
raw_boxes_ = edge_boxes.get_windows([img_id])[0][0:top_num, :]
raw_boxes = np.zeros(raw_boxes_.shape)
raw_boxes[:, 0] = raw_boxes_[:, 1]
raw_boxes[:, 1] = raw_boxes_[:, 0]
raw_boxes[:, 2] = raw_boxes_[:, 3] - raw_boxes_[:, 1] + 1
raw_boxes[:, 3] = raw_boxes_[:, 2] - raw_boxes_[:, 0] + 1
edge_boxes = np.zeros((0, 4))
edge_boxes = np.concatenate((edge_boxes, raw_boxes[:, 0:4]))
return edge_boxes
def demo(net, image_name, classes):
"""Detect object classes in an image using pre-computed object proposals."""
# Load the demo image
im_file = os.path.join(cfg.ROOT_DIR, 'data', 'demo', image_name + '.jpg')
im = cv2.imread(im_file)
# Compute EdgeBoxes proposals
image_filenames = [ im_file ]
initial_boxes = get_windows(image_filenames);
# Get just the boxes, without the EdgeBoxes confidence scores
obj_proposals = (initial_boxes[0])[:, [0,1,2,3]]
# Detect all object classes and regress object bounds
timer = Timer()
timer.tic()
scores, boxes = im_detect(net, im, obj_proposals)
timer.toc()
print ('Detection took {:.3f}s for '
'{:d} object proposals').format(timer.total_time, boxes.shape[0])
# Visualize detections for each class
CONF_THRESH = 0.8
NMS_THRESH = 0.3
for cls in classes:
cls_ind = CLASSES.index(cls)
cls_boxes = boxes[:, 4*cls_ind:4*(cls_ind + 1)]
cls_scores = scores[:, cls_ind]
dets = np.hstack((cls_boxes,
cls_scores[:, np.newaxis])).astype(np.float32)
keep = nms(dets, NMS_THRESH)
dets = dets[keep, :]
print 'All {} detections with p({} | box) >= {:.1f}'.format(cls, cls,
CONF_THRESH)
vis_detections(im, cls, dets, thresh=CONF_THRESH)
import edge_boxes as edges windows = edges.get_windows(['/home/zawlin/data/indoor.jpg']) print windows
# Why? Of the 15 seconds, something like 14.5 of them are "waiting for Matlab to load". Only
# 0.5 of them are doing work. If you have to loop over 400 images, waiting 14.5 seconds per image
# (for Matlab to load) adds up to something like 100 minutes. If you call the edge_boxes function
# once (with paths to all 400 images), it only takes 3 minutes to run on ecegrid.
# SO... again my advice: develop your code by passing in a single image (since waiting 15 seconds
# is better than waiting 3 minutes). Then, when you're ready to run your code on the entire
# directory of images, tweak your code to call edge_boxes once.
# NOTE: When I had finished coding and was ready to run my code on the entire directory of images,
# it took ~15 minutes from beginning to end. Each directory of images (4 total) took ~3 minutes
# for edge_boxes to run and ~1.5 minutes for the discovery, sorting, and saving of pos/neg images.
# What can you do while you wait 15 minutes for your code to complete? Start working on phase2!
#################################################################################################
#imageFileNames = []
#imageFileNames.append(imageDir + "/n02119789_122.JPEG")
#imageFileNames.append(imageDir + "/n02119789_257.JPEG")
windows = edge_boxes.get_windows(imageFileNames)
#print(windows[1])
pos_num = 1
neg_num = 1
for i in range(len(imageFileNames)):
orig_image = cv2.imread(imageFileNames[i])
neg = 0 # count how many negative proposal
pos = 0 # count how many positive proposal
boxnum = 0
loop_num = 0
tree = ET.parse(annotationFileName[i]) # ground truth
objs_ground = tree.findall('object') # ground truth
#iou_his = []
while (pos < 5 and neg < 5): # for each image
loop_num = loop_num + 1
iou_his = []
num = 0
w = []
ka = "/home/min/a/prachapu/ee570/hw3/"
while True:
line = f.readline()
if not line:
break
l = line.split()
pa = l[0][2:]
l[0] = ka + pa
# print line, l
# print l
print(l[0])
w = []
w.append(l[0])
windows = edge_boxes.get_windows(w)
windows = numpy.array(windows)
image = cv2.imread(l[0])
# cv2.imshow('image',image)
# cv2.waitKey(0)
#print windows
prop_ctr = 0
best = -1
result = []
store = []
for u in xrange(0, 4):
result.append(0)
store.append([])
for proposals in windows:
for boxes in proposals:
p = []