How to infer the SGs on my own dataset?

[ZihaoZheng98]

No description provided.

[jinbae]

For inference, you have to prepare your data loader.
Please refer to these lines:

egtr/evaluate_egtr.py

Lines 178 to 204 in b3f7b41

	# Dataset
	if "visual_genome" in args.data_path:
	test_dataset = VGDataset(
	data_folder=args.data_path,
	feature_extractor=feature_extractor,
	split=args.split,
	num_object_queries=args.num_queries,
	)
	id2label = {
	k - 1: v["name"] for k, v in test_dataset.coco.cats.items()
	} # 0 ~ 149
	coco_evaluator = CocoEvaluator(
	test_dataset.coco, ["bbox"]
	) # initialize evaluator with ground truths
	oi_evaluator = None
	elif "open-image" in args.data_path:
	test_dataset = OIDataset(
	data_folder=args.data_path,
	feature_extractor=feature_extractor,
	split=args.split,
	num_object_queries=args.num_queries,
	)
	id2label = test_dataset.classes_to_ind # 0 ~ 600
	oi_evaluator = OIEvaluator(
	test_dataset.rel_categories, test_dataset.ind_to_classes
	)
	coco_evaluator = None

You do not need to develop your evaluator, but you need your data processing code.
Please refer to the data folder.