在图像上标记目标的矩形(bounding box, bbox)。常见的标注格式为Pascal VOC、COCO、YOLO

Pascal VOC

bbox：[x_min, y_min, x_max, y_max]

格式：左上右下

COCO

bbox：[x_min, ymin, width, height]

格式：左上宽高

YOLO

bbox [x_center, y_center, width, height]

并进行数据规范化(normalized)

格式：中心坐标，宽高

Pasic VOC 转 YOLO

    def convert_box(size, box):
        # Convert VOC box to YOLO xywh box
        dw = 1. / size[0]
        dh = 1. / size[1]

        return ((box[0] + box[1]) / 2.0 * dw, (box[2] + box[3]) / 2.0 * dh , (box[1] - box[0]) * dw, (box[3] - box[2]) * * dh)

COCO 转 YOLO

    def convert_box(size, box):
        # Convert COCO box to YOLO xywh box
        dw = 1. / size[0]
        dh = 1. / size[1]

        return (box[0] + box[2] / 2) * dw, (box[1] + box[3] / 2) * dh, box[2] * dw, box[3] * dh

YOLO Decode

def xywhn2xyxy(x, w=640, h=640, padw=0, padh=0):
    # 将yolo格式的box直接读取
    # Convert nx4 boxes from [x, y, w, h] normalized to [x1, y1, x2, y2] where xy1=top-left, xy2=bottom-right
    y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x)
    y[:, 0] = w * (x[:, 0] - x[:, 2] / 2) + padw  # top left x
    y[:, 1] = h * (x[:, 1] - x[:, 3] / 2) + padh  # top left y
    y[:, 2] = w * (x[:, 0] + x[:, 2] / 2) + padw  # bottom right x
    y[:, 3] = h * (x[:, 1] + x[:, 3] / 2) + padh  # bottom right y
    return y