def __init__(self, imgset, classwise, ann_file, img_prefix, *args,
**kwargs):
self.imgset = imgset
if classwise:
HRSCDataset.CLASSES = bt.get_classes('hrsc_cls')
else:
HRSCDataset.CLASSES = bt.get_classes('hrsc')
super(HRSCDataset, self).__init__(*args,
ann_file=ann_file,
img_prefix=img_prefix,
**kwargs)
def get_classes(cls, classes=None):
if classes is None:
cls.custom_classes = False
return cls.CLASSES
cls.custom_classes = True
return bt.get_classes(classes)
def main():
args = parse_args()
print(f'{args.load_type} loading!')
load_func = getattr(bt.datasets, 'load_' + args.load_type)
contents, classes = load_func(img_dir=args.img_dir,
ann_dir=args.ann_dir,
classes=args.classes,
nproc=args.load_nproc)
if args.prior_annfile is not None:
prior, _ = bt.load_pkl(args.prior_annfile, classes=classes)
bt.merge_prior_contents(contents, prior, merge_type=args.merge_type)
shown_names = classes if args.shown_names is None \
else bt.get_classes(args.shown_names)
assert len(shown_names) == len(classes)
if isinstance(args.ids, (list, type(None))):
ids = args.ids
elif isinstance(args.ids, str):
if osp.isfile(args.ids):
with open(args.ids, 'r') as f:
ids = [l.strip() for l in f]
else:
ids = args.ids.split('|')
else:
raise TypeError('Wrong base_json input in `ids`')
tasks, max_label = [], 0
for content in contents:
if ids is not None and content['id'] not in ids:
pass
imgpath = osp.join(args.img_dir, content['filename'])
out_file = osp.join(args.save_dir, content['filename']) \
if args.save_dir else None
if 'ann' in content:
ann = content['ann']
bboxes = ann['bboxes']
labels = ann['labels']
scores = ann.get('scores', None)
else:
bboxes = np.zeros((0, 4), dtype=np.float)
labels = np.zeros((0, ), dtype=np.int)
scores = None
if (scores is not None) and (args.score_thr > 0):
bboxes = bboxes[scores > args.score_thr]
labels = labels[scores > args.score_thr]
scores = scores[scores > args.score_thr]
if args.skip_empty and bboxes.size == 0:
continue
if labels.size > 0:
max_label = max(max_label, labels.max())
tasks.append((imgpath, out_file, bboxes, labels, scores))
if args.colors == 'random':
args.colors = bt.random_colors(max_label + 1)
if args.random_vis:
shuffle(tasks)
if args.save_dir and (not osp.exists(args.save_dir)):
os.makedirs(args.save_dir)
if args.show_off:
plt.switch_backend('Agg')
manager = Manager()
_vis_func = partial(single_vis,
btype=args.shown_btype,
class_names=shown_names,
colors=args.colors,
thickness=args.thickness,
text_off=args.text_off,
font_size=args.font_size,
show_off=args.show_off,
wait_time=args.wait_time,
lock=manager.Lock(),
prog=manager.Value('i', 0),
total=len(tasks))
if args.show_off and args.vis_nproc > 1:
pool = Pool(args.vis_nproc)
pool.map(_vis_func, tasks)
pool.close()
else:
list(map(_vis_func, tasks))
if args.save_dir:
arg_dict = vars(args)
arg_dict.pop('base_json', None)
with open(osp.join(args.save_dir, 'vis_config.json'), 'w') as f:
json.dump(arg_dict, f, indent=4)
class DIORDataset(CustomDataset):
CLASSES = bt.get_classes('dior')
def __init__(self,
xmltype,
imgset,
ann_file,
img_prefix,
*args,
**kwargs):
assert xmltype in ['hbb', 'obb']
self.xmltype = xmltype
self.imgset = imgset
super(DIORDataset, self).__init__(*args,
ann_file=ann_file,
img_prefix=img_prefix,
**kwargs)
@classmethod
def get_classes(cls, classes=None):
if classes is None:
cls.custom_classes = False
return cls.CLASSES
cls.custom_classes = True
return bt.get_classes(classes)
def load_annotations(self, ann_file):
contents, _ = bt.load_dior(
img_dir=self.img_prefix,
ann_dir=ann_file,
classes=self.CLASSES,
xmltype=self.xmltype)
if self.imgset is not None:
contents = bt.split_imgset(contents, self.imgset)
return contents
def pre_pipeline(self, results):
results['cls'] = self.CLASSES
super().pre_pipeline(results)
def format_results(self, results, save_dir=None, **kwargs):
assert len(results) == len(self.data_infos)
contents = []
for result, data_info in zip(results, self.data_infos):
info = copy.deepcopy(data_info)
info.pop('ann')
ann, bboxes, labels, scores = dict(), list(), list(), list()
for i, dets in enumerate(result):
bboxes.append(dets[:, :-1])
scores.append(dets[:, -1])
labels.append(np.zeros((dets.shape[0], ), dtype=np.int) + i)
ann['bboxes'] = np.concatenate(bboxes, axis=0)
ann['labels'] = np.concatenate(labels, axis=0)
ann['scores'] = np.concatenate(scores, axis=0)
info['ann'] = ann
contents.append(info)
if save_dir is not None:
bt.save_pkl(save_dir, contents, self.CLASSES)
return contents
def evaluate(self,
results,
metric='mAP',
logger=None,
iou_thr=0.5,
scale_ranges=None,
use_07_metric=True,
proposal_nums=(100, 300, 1000)):
if not isinstance(metric, str):
assert len(metric) == 1
metric = metric[0]
allowed_metrics = ['mAP', 'recall']
if metric not in allowed_metrics:
raise KeyError(f'metric {metric} is not supported')
annotations = [self.get_ann_info(i) for i in range(len(self))]
eval_results = {}
if metric == 'mAP':
assert isinstance(iou_thr, float)
mean_ap, _ = eval_arb_map(
results,
annotations,
scale_ranges=scale_ranges,
iou_thr=iou_thr,
use_07_metric=use_07_metric,
dataset=self.CLASSES,
logger=logger)
eval_results['mAP'] = mean_ap
elif metric == 'recall':
gt_bboxes = [ann['bboxes'] for ann in annotations]
if isinstance(iou_thr, float):
iou_thr = [iou_thr]
recalls = eval_arb_recalls(
gt_bboxes, results, True, proposal_nums, iou_thr, logger=logger)
for i, num in enumerate(proposal_nums):
for j, iou in enumerate(iou_thr):
eval_results[f'recall@{num}@{iou}'] = recalls[i, j]
if recalls.shape[1] > 1:
ar = recalls.mean(axis=1)
for i, num in enumerate(proposal_nums):
eval_results[f'AR@{num}'] = ar[i]
return eval_results
def print_map_summary(mean_ap,
results,
dataset=None,
scale_ranges=None,
logger=None):
"""Print mAP and results of each class.
A table will be printed to show the gts/dets/recall/AP of each class and
the mAP.
Args:
mean_ap (float): Calculated from `eval_map()`.
results (list[dict]): Calculated from `eval_map()`.
dataset (list[str] | str | None): Dataset name or dataset classes.
scale_ranges (list[tuple] | None): Range of scales to be evaluated.
logger (logging.Logger | str | None): The way to print the mAP
summary. See `mmdet.utils.print_log()` for details. Default: None.
"""
if logger == 'silent':
return
if isinstance(results[0]['ap'], np.ndarray):
num_scales = len(results[0]['ap'])
else:
num_scales = 1
if scale_ranges is not None:
assert len(scale_ranges) == num_scales
num_classes = len(results)
recalls = np.zeros((num_scales, num_classes), dtype=np.float32)
aps = np.zeros((num_scales, num_classes), dtype=np.float32)
num_gts = np.zeros((num_scales, num_classes), dtype=int)
for i, cls_result in enumerate(results):
if cls_result['recall'].size > 0:
recalls[:, i] = np.array(cls_result['recall'], ndmin=2)[:, -1]
aps[:, i] = cls_result['ap']
num_gts[:, i] = cls_result['num_gts']
if dataset is None:
label_names = [str(i) for i in range(num_classes)]
elif mmcv.is_str(dataset):
label_names = bt.get_classes(dataset)
else:
label_names = dataset
if not isinstance(mean_ap, list):
mean_ap = [mean_ap]
header = ['class', 'gts', 'dets', 'recall', 'ap']
for i in range(num_scales):
if scale_ranges is not None:
print_log(f'Scale range {scale_ranges[i]}', logger=logger)
table_data = [header]
for j in range(num_classes):
row_data = [
label_names[j], num_gts[i, j], results[j]['num_dets'],
f'{recalls[i, j]:.4f}', f'{aps[i, j]:.4f}'
]
table_data.append(row_data)
table_data.append(['mAP', '', '', '', f'{mean_ap[i]:.4f}'])
table = AsciiTable(table_data)
table.inner_footing_row_border = True
print_log('\n' + table.table, logger=logger)
class RCTW17Dataset(CustomDataset):
CLASSES = bt.get_classes('RCTW-17')
def __init__(self, ann_file, img_prefix, imgset=None, *args, **kwargs):
self.imgset = imgset
super(RCTW17Dataset, self).__init__(*args,
ann_file=ann_file,
img_prefix=img_prefix,
**kwargs)
@classmethod
def get_classes(cls, classes=None):
if classes is None:
cls.custom_classes = False
return cls.CLASSES
cls.custom_classes = True
return bt.get_classes(classes)
def load_annotations(self, ann_file):
contents, _ = bt.load_rctw_17(img_dir=self.img_prefix,
ann_dir=ann_file,
classes=self.CLASSES)
if self.imgset is not None:
contents = bt.split_imgset(contents, self.imgset)
return contents
def pre_pipeline(self, results):
results['cls'] = self.CLASSES
super().pre_pipeline(results)
def format_results(self, results, save_dir=None, **kwargs):
assert len(results) == len(self.data_infos)
id_list = [info['id'] for info in self.data_infos]
if save_dir is not None:
bt.save_rctw_17(save_dir, id_list, results)
return results
def evaluate(self,
results,
metric='mAP',
logger=None,
iou_thr=0.5,
scale_ranges=None,
use_07_metric=True,
proposal_nums=(100, 300, 1000)):
if not isinstance(metric, str):
assert len(metric) == 1
metric = metric[0]
allowed_metrics = ['mAP', 'recall']
if metric not in allowed_metrics:
raise KeyError(f'metric {metric} is not supported')
annotations = [self.get_ann_info(i) for i in range(len(self))]
eval_results = {}
if metric == 'mAP':
assert isinstance(iou_thr, float)
mean_ap, _ = eval_arb_map(results,
annotations,
scale_ranges=scale_ranges,
iou_thr=iou_thr,
use_07_metric=use_07_metric,
dataset=self.CLASSES,
logger=logger)
eval_results['mAP'] = mean_ap
elif metric == 'recall':
gt_bboxes = [ann['bboxes'] for ann in annotations]
if isinstance(iou_thr, float):
iou_thr = [iou_thr]
recalls = eval_arb_recalls(gt_bboxes,
results,
True,
proposal_nums,
iou_thr,
logger=logger)
for i, num in enumerate(proposal_nums):
for j, iou in enumerate(iou_thr):
eval_results[f'recall@{num}@{iou}'] = recalls[i, j]
if recalls.shape[1] > 1:
ar = recalls.mean(axis=1)
for i, num in enumerate(proposal_nums):
eval_results[f'AR@{num}'] = ar[i]
return eval_results