|
|
import datasets |
|
|
import os |
|
|
import numpy as np |
|
|
|
|
|
|
|
|
|
|
|
def parse_bbox_file(filepath): |
|
|
"""Đọc file bbox và chuyển đổi thành list các số thực.""" |
|
|
if not os.path.exists(filepath): |
|
|
return [] |
|
|
with open(filepath, 'r', encoding='utf-8') as f: |
|
|
line = f.read().strip() |
|
|
if not line: |
|
|
return [] |
|
|
return [float(coord) for coord in line.split(',')] |
|
|
|
|
|
def parse_points_file(filepath): |
|
|
"""Đọc file points và chuyển đổi thành list của các list số nguyên.""" |
|
|
if not os.path.exists(filepath): |
|
|
return [] |
|
|
points = [] |
|
|
with open(filepath, 'r', encoding='utf-8') as f: |
|
|
for line in f: |
|
|
line = line.strip() |
|
|
if not line: |
|
|
continue |
|
|
coords = [int(p.strip()) for p in line.split(',')] |
|
|
points.append(coords) |
|
|
return points |
|
|
|
|
|
|
|
|
|
|
|
class ORiDaFactualCounterfactual(datasets.GeneratorBasedBuilder): |
|
|
"""Dataset loader cho bộ dữ liệu ORiDa Factual/Counterfactual.""" |
|
|
|
|
|
def _info(self): |
|
|
|
|
|
return datasets.DatasetInfo( |
|
|
description="Bộ dữ liệu ORiDa chứa các mẫu factual và counterfactual.", |
|
|
features=datasets.Features({ |
|
|
'group_id': datasets.Value("string"), |
|
|
'sample_id': datasets.Value("string"), |
|
|
'type': datasets.ClassLabel(names=['factual_only', 'factual_counterfactual']), |
|
|
'image': datasets.Image(), |
|
|
'ground_truth': datasets.Image(), |
|
|
'annotations': { |
|
|
'bbox': datasets.Sequence(datasets.Value("float32")), |
|
|
'points': datasets.Sequence(datasets.Sequence(datasets.Value("int32"))), |
|
|
'masks': datasets.Image(), |
|
|
'results_vis': datasets.Image(), |
|
|
} |
|
|
}), |
|
|
) |
|
|
|
|
|
def _split_generators(self, dl_manager): |
|
|
|
|
|
train_path = dl_manager.extract("train.zip") |
|
|
validation_path = dl_manager.extract("validation.zip") |
|
|
|
|
|
|
|
|
return [ |
|
|
datasets.SplitGenerator( |
|
|
name=datasets.Split.TRAIN, |
|
|
gen_kwargs={"filepath": os.path.join(train_path, "train")} |
|
|
), |
|
|
datasets.SplitGenerator( |
|
|
name=datasets.Split.VALIDATION, |
|
|
gen_kwargs={"filepath": os.path.join(validation_path, "validation")} |
|
|
), |
|
|
] |
|
|
|
|
|
def _generate_examples(self, filepath): |
|
|
"""Hàm này sẽ duyệt qua thư mục và 'yield' từng mẫu dữ liệu.""" |
|
|
key = 0 |
|
|
for group_id in os.listdir(filepath): |
|
|
group_path = os.path.join(filepath, group_id) |
|
|
if not os.path.isdir(group_path): continue |
|
|
|
|
|
ground_truth_path = os.path.join(group_path, f"{group_id}.png") |
|
|
if not os.path.exists(ground_truth_path): continue |
|
|
|
|
|
for type_name in ["factual_only", "factual_counterfactual"]: |
|
|
type_path = os.path.join(group_path, type_name) |
|
|
if not os.path.isdir(type_path): continue |
|
|
|
|
|
for sample_id in os.listdir(type_path): |
|
|
sample_path = os.path.join(type_path, sample_id) |
|
|
if not os.path.isdir(sample_path): continue |
|
|
|
|
|
images_path = os.path.join(sample_path, "images") |
|
|
annotations_path = os.path.join(sample_path, "annotations") |
|
|
if not os.path.isdir(images_path): continue |
|
|
|
|
|
for image_filename in os.listdir(images_path): |
|
|
if not image_filename.lower().endswith(('.png', '.jpg', '.jpeg')): continue |
|
|
|
|
|
image_path = os.path.join(images_path, image_filename) |
|
|
base_image_name = os.path.splitext(image_filename)[0] |
|
|
|
|
|
|
|
|
bbox_file = os.path.join(annotations_path, "bbox", f"{base_image_name}.txt") |
|
|
points_file = os.path.join(annotations_path, "points", f"{base_image_name}.txt") |
|
|
|
|
|
mask_file = os.path.join(annotations_path, "masks", f"{base_image_name}.jpg") |
|
|
results_vis_file = os.path.join(annotations_path, "results_vis", f"{base_image_name}.jpg") |
|
|
|
|
|
|
|
|
bbox_data = parse_bbox_file(bbox_file) |
|
|
points_data = parse_points_file(points_file) |
|
|
|
|
|
|
|
|
if not os.path.exists(mask_file) or not os.path.exists(results_vis_file): |
|
|
continue |
|
|
|
|
|
yield key, { |
|
|
"group_id": group_id, |
|
|
"sample_id": sample_id, |
|
|
"type": type_name, |
|
|
"image": image_path, |
|
|
"ground_truth": ground_truth_path, |
|
|
"annotations": { |
|
|
"bbox": bbox_data, |
|
|
"points": points_data, |
|
|
"masks": mask_file, |
|
|
"results_vis": results_vis_file, |
|
|
} |
|
|
} |
|
|
key += 1 |