def convert_to_frame_predictions(detections_dir):
result = {}
detection_files = helper.get_files(detections_dir)
for file_name in detection_files:
read_file = os.path.join(detections_dir, file_name)
with open(read_file, 'r') as rf:
file_info = {}
width, height = [int(i) for i in rf.readline().split()
] # first line is image size
file_info["image_height"] = height
file_info["image_width"] = width
file_info["detections"] = []
for line in rf:
det = helper.Detection(line)
detection = {}
detection["bbox"] = [det.x1, det.y1, det.x2, det.y2]
detection["prob"] = det.score
detection["class"] = str(det.class_id) + " " + det.class_name
file_info["detections"].append(detection)
result[file_name] = file_info
return result
def convert_to_kitti(detection_files, detections_dir, target_dir,
model_dataset_type):
for file_name in detection_files:
read_file = os.path.join(detections_dir, file_name)
write_file = os.path.join(target_dir, file_name)
with open(read_file, 'r') as rf, open(write_file, 'w') as wf:
rf.readline() # first line is image size
for line in rf:
det = helper.Detection(line)
res = detection_to_kitti_string(det, model_dataset_type)
if (res):
wf.write(res)
return target_dir
def convert_to_coco(detection_files, detections_dir, target_dir, dataset_type,
model_dataset_type):
res_array = []
for file_name in detection_files:
read_file = os.path.join(detections_dir, file_name)
file_id = helper.filename_to_id(file_name, dataset_type)
with open(read_file, 'r') as rf:
rf.readline() # first line is image size
for line in rf:
det = helper.Detection(line)
res = detection_to_coco_object(det, model_dataset_type,
file_id)
if (res):
res_array.append(res)
results_file = os.path.join(target_dir, 'coco_results.json')
with open(results_file, 'w') as f:
f.write(json.dumps(res_array, indent=2, sort_keys=False))
return results_file
def fill_detection(detection, file_name):
with open(file_name, 'r') as f:
size = f.readline().split()
x = float(size[0])
y = float(size[1])
cnt = 0
for line in f:
det = ck_utils.Detection(line)
cnt += 1
detection['detection_boxes'] = np.append(detection['detection_boxes'], [[det.x1/x, det.y1/y, det.x2/x, det.y2/y]], 0)
detection['detection_classes'] = np.append(detection['detection_classes'], det.class_id)
detection['detection_scores'] = np.append(detection['detection_scores'], det.score)
# Detections dict can't be empty even there is not detection for this image
if cnt == 0:
detection['detection_boxes'] = np.array([[0.0, 0.0, 0.0, 0.0]])
detection['detection_classes'] = np.array([0])
detection['detection_scores'] = np.array([0.0])
return detection
def load_detections(file_path):
boxes = []
scores = []
classes = []
with open(file_path, 'r') as f:
f.readline() # first line is image size
for line in f:
det = ck_utils.Detection(line)
boxes.append([det.x1, det.y1, det.x2, det.y2])
scores.append(det.score)
classes.append(det.class_id)
# Detections dict can't be empty even there is not detection for this image
if not boxes:
boxes = [[0, 0, 0, 0]]
scores = [0]
classes = [0]
return {
det_field.detection_boxes: np.array(boxes),
det_field.detection_scores: np.array(scores),
det_field.detection_classes: np.array(classes)
}