def add_2d_detection(img,
dets,
expected_cls,
show_txt=True,
center_thresh=0.5):
right_label = 0
wrong_labels = 0
for cat in dets:
for i in range(len(dets[cat])):
if dets[cat][i, -1] > center_thresh:
if cat - 1 == expected_cls:
right_label += 1
else:
wrong_labels += 1
bbox = dets[cat][i, :4]
add_bbox(
img,
[int(bbox[0]),
int(bbox[1]),
int(bbox[2]),
int(bbox[3])],
cat - 1, ['ov', 'mif'],
dets[cat][i, -1],
show_txt=show_txt)
if right_label > 0 and wrong_labels == 0:
score = 1
elif right_label == 0 and wrong_labels > 0:
score = 2
else:
score = 0
if right_label + wrong_labels > 0:
score = 3
return img, score
for box, score, label in zip(boxes[0], scores[0], labels[0]):
# scores are sorted so we can break
if score < 0.5:
break
# color = label_color(label)
if (label == 0
and 'mif' in pth.lower()) or (label == 1
and 'ov' in pth.lower()):
right_label.append(label)
else:
wrong_labels.append(label)
b = box.astype(int)
# draw_box(draw, b, color=[(255, 0, 0), (0, 255, 0)][label])
draw = add_bbox(draw, b, label, labels_to_names, score)
caption = "{} {:.3f}".format(labels_to_names[label], score)
# draw_caption(draw, b, caption)
if len(wrong_labels) == 0 and len(right_label) > 0:
flag = 'correct'
elif len(wrong_labels) > 0 and len(right_label) > 0:
flag = 'mixed'
elif len(wrong_labels) == 0 and len(right_label) == 0:
flag = 'not_found'
else:
flag = 'wrong'
img = Image.fromarray(draw)
img.save(
os.path.join(
# scores are sorted so we can break
if score < 0.5:
break
b = box.astype(int)
minimum = (float('inf'), 0)
with torch.no_grad():
target_image_ori = pad(draw[b[1]:b[3], b[0]:b[2]], b)
target_image_ori = Image.fromarray(target_image_ori[..., ::-1])
target_image = transform(target_image_ori)
x = torch.zeros((1, 3, 224, 224))
x[0] = target_image
target_features = model.model._forward_impl(x.cuda())
for query_folder in os.listdir(query_path):
for query_image_path in os.listdir(os.path.join(query_path, query_folder)):
query = os.path.join(query_path, query_folder, query_image_path)
cache_dict, query_features = memory_cache(cache_dict, model.model, query, os.path.join(cache_path, query_folder, query_image_path + '.pth'), transform)
y = LSHash.euclidean_dist(target_features.cpu().numpy()[0], query_features.cpu().numpy()[0])
if y < minimum[0]:
minimum = (y, query_folder)
if minimum[0] > 1:
minimum = (minimum[0], 'obj')
else:
if minimum[1] in names_to_labels:
label = names_to_labels[minimum[1]]
draw = add_bbox(draw, b, label, labels_to_names, score)
else:
draw = add_bbox(draw, b, 0, [minimum[1]], score)
os.makedirs(dst, exist_ok=True)
cv2.imwrite(os.path.join(dst, i), draw)
color = [(0, 0, 255), (255, 0, 0)
][classes[i] == int('mif' in d["filename"].lower())]
if 'mif' in d["filename"].lower():
if classes[i] == 1:
color = (255, 0, 0)
else:
color = (0, 0, 255)
else:
if classes[i] == 0:
color = (255, 0, 0)
else:
color = (0, 0, 255)
im = add_bbox(im,
bboxes[i].astype('uint16'),
classes[i], ['ov', 'mif'],
scores[i],
show_txt=True,
color=color)
for i in range(len(d['object'])):
box = [
d['object'][i]['xmin'], d['object'][i]['ymin'],
d['object'][i]['xmax'], d['object'][i]['ymax']
]
im = add_bbox(im,
box,
2, ['ov', 'mif', 'obj'],
show_txt=False,
color=(0, 200, 0))
cv2.imwrite(
os.path.join(
import cv2
import csv
from retinanet.preprocessing.csv_generator import _read_annotations, _open_for_csv, _read_classes
from boxutils import add_bbox
if __name__ == '__main__':
with _open_for_csv(
'/home/palm/PycharmProjects/algea/dataset/classes') as file:
classes = _read_classes(csv.reader(file, delimiter=','))
with _open_for_csv(
'/home/palm/PycharmProjects/algea/dataset/train_annotations'
) as file:
train_image_data = _read_annotations(csv.reader(file, delimiter=','),
classes)
with _open_for_csv(
'/home/palm/PycharmProjects/algea/dataset/test_annotations'
) as file:
test_image_data = _read_annotations(csv.reader(file, delimiter=','),
classes)
for k in train_image_data:
anns = train_image_data[k]
image = cv2.imread(k)
for ann in anns:
bbox = (ann['x1'], ann['y1'], ann['x2'], ann['y2'])
image = add_bbox(image, bbox, classes[ann['class']], ['mif', 'ov'])
cv2.imwrite(
f'/home/palm/PycharmProjects/algea/dataset/plotted/{k.split("/")[-1]}',
image)