def show_result_rbox(img,
detections,
class_names,
scale=1.0,
threshold=0.2,
colormap=None,
show_label=False):
assert isinstance(class_names, (tuple, list))
if colormap:
assert len(class_names) == len(colormap)
img = mmcv.imread(img)
color_white = (255, 255, 255)
for j, name in enumerate(class_names):
if colormap:
color = colormap[j]
else:
color = (random.randint(0, 256), random.randint(0, 256),
random.randint(0, 256))
try:
dets = detections[j]
except:
# pdb.set_trace()
print('no result')
return img # TODO don't debug
# import ipdb;ipdb.set_trace()
for det in dets:
score = det[-1]
det = rotated_box_to_poly_single(det[:-1])
bbox = det[:8] * scale
if score < threshold:
continue
bbox = list(map(int, bbox))
for i in range(3):
cv2.line(img, (bbox[i * 2], bbox[i * 2 + 1]),
(bbox[(i + 1) * 2], bbox[(i + 1) * 2 + 1]),
color=color,
thickness=2,
lineType=cv2.LINE_AA)
cv2.line(img, (bbox[6], bbox[7]), (bbox[0], bbox[1]),
color=color,
thickness=2,
lineType=cv2.LINE_AA)
if show_label:
cv2.putText(img,
'%s %.3f' % (class_names[j], score),
(bbox[0], bbox[1] + 10),
color=color_white,
fontFace=cv2.FONT_HERSHEY_COMPLEX,
fontScale=0.5)
return img
def parse_gt(filename):
objects = []
tree = ET.parse(filename)
root = tree.getroot()
for obj in root.findall('HRSC_Objects')[0].findall('HRSC_Object'):
object_struct = {}
object_struct['name'] = 'ship'
object_struct['difficult'] = int(obj.find('difficult').text)
bbox = []
for key in ['mbox_cx', 'mbox_cy', 'mbox_w', 'mbox_h', 'mbox_ang']:
bbox.append(obj.find(key).text)
# Coordinates may be float type
cx, cy, w, h, a = list(map(float, bbox))
bbox = [cx, cy, w, h, a]
poly = rotated_box_to_poly_single(bbox)
object_struct['bbox'] = poly.tolist()
objects.append(object_struct)
return objects
def show_result_rbox(img,
detections,
class_names,
scale=1.0,
threshold=0.2,
colormap=None,
show_label=False):
assert isinstance(class_names, (tuple, list))
if colormap:
assert len(class_names) == len(colormap)
img = mmcv.imread(img)
color_white = (255, 255, 255)
for dets, name in zip(detections['bboxes'], detections['labels']):
if colormap:
color = colormap[name-1]
else:
color = (random.randint(0, 256), random.randint(0, 256), random.randint(0, 256))
for det in dets.reshape(1, -1):
cv2.circle(img, (det[0], det[1]), radius=2, color=color, thickness=2)
cv2.putText(img, '%.3f' % (det[4]), (det[0], det[1]),
color=color_white, fontFace=cv2.FONT_HERSHEY_COMPLEX, fontScale=0.5)
score = 1.0
det = rotated_box_to_poly_single(det)
bbox = det[:8] * scale
if score < threshold:
continue
bbox = list(map(int, bbox))
for i in range(3):
cv2.line(img, (bbox[i * 2], bbox[i * 2 + 1]), (bbox[(i + 1) * 2], bbox[(i + 1) * 2 + 1]), color=color,
thickness=2, lineType=cv2.LINE_AA)
cv2.line(img, (bbox[6], bbox[7]), (bbox[0], bbox[1]), color=color, thickness=2, lineType=cv2.LINE_AA)
if show_label:
cv2.putText(img, '%s %.3f' % (class_names[j], score), (bbox[0], bbox[1] + 10),
color=color_white, fontFace=cv2.FONT_HERSHEY_COMPLEX, fontScale=0.5)
return img
def result_to_txt(self, results, results_path):
img_names = [img_info['id'] for img_info in self.img_infos]
assert len(results) == len(img_names), 'len(results) != len(img_names)'
for classname in self.CLASSES:
f_out = open(osp.join(results_path, classname + '.txt'), 'w')
print(classname + '.txt')
# per result represent one image
for img_id, result in enumerate(results):
for class_id, bboxes in enumerate(result):
if self.CLASSES[class_id] != classname:
continue
if bboxes.size != 0:
for bbox in bboxes:
score = bbox[5]
bbox = rotated_box_to_poly_single(bbox[:5])
temp_txt = '{} {:.4f} {:.4f} {:.4f} {:.4f} {:.4f} {:.4f} {:.4f} {:.4f} {:.4f}\n'.format(
osp.splitext(img_names[img_id])[0], score, bbox[0], bbox[1], bbox[2], bbox[3], bbox[4],
bbox[5], bbox[6], bbox[7])
f_out.write(temp_txt)
f_out.close()