Dataset EIPH_WSI about objectdetection HOT 2 CLOSED

CATEGORIES = []
for i in range(5):
    CATEGORIES += {
        'id': str(i),
        'name': str(i),
        'supercategory': 'Hemosiderophages',
    },

coco_output = {
    # "info": INFO,
    #  "licenses": LICENSES,
    "categories": CATEGORIES,
    "images": [],
    "annotations": []
}

image_id = 1
annotation_id = 1

path = Path('/data/Datasets/)

# for each file to convert into the coco format
for file in tqdm(train_files):

    # get boxes and labels in the format [[x1,x2,w,h],   [x1,x2,w,h]], [1,2]
    boxes, labels = get_y_func(file)
    boxes = np.array(boxes)
    labels = np.array(labels)
    classes = list(set(labels))

    for image, y in zip(x_batch, y_batch):
        boxes, labels = y

        name = file.file.stem + "_" + str(image_id) + ".png"

        image_info = {
            "id": image_id,
            "file_name": name,
            "width": image.shape[0],
            "height": image.shape[1],
        }

        coco_output["images"].append(image_info)

        for box, label in zip(boxes, labels):
            annotation_info = {
                "id": annotation_id,
                "image_id": image_id,
                "category_id": str(label),
                "bbox": [int(box[0]), int(box[1]), int(box[2]-box[0]), int(box[3]-box[1])]
            }
            coco_output["annotations"].append(annotation_info)
            annotation_id = annotation_id + 1

        image_id += 1

with open(str(path/'WIPH_WSI_Coco.json'), 'w') as output_json_file:
    json.dump(coco_output, output_json_file)