替换cython_bbox库中bbox_ious

说明一下问题：目标追踪代码里往往用到cython_bbox中的bbox_ious。但是该库需要用到 VC++，按照有些麻烦。于是采用直接替换该方法，用纯代码实现，无需调用库。

File “src\cython_bbox.pyx”, line 17, in cython_bbox.bbox_overlaps
ValueError: Buffer dtype mismatch, expected ‘DTYPE_t’ but got ‘float’

代码来源：
PaddleDetection-release-2.7\build\lib\ppdet\data\transform\atss_assigner.py

def bbox_ious(bboxes1, bboxes2, mode='iou', is_aligned=False, eps=1e-6):
    """Calculate overlap between two set of bboxes.
    If ``is_aligned `` is ``False``, then calculate the overlaps between each
    bbox of bboxes1 and bboxes2, otherwise the overlaps between each aligned
    pair of bboxes1 and bboxes2.
    Args:
        bboxes1 (Tensor): shape (B, m, 4) in <x1, y1, x2, y2> format or empty.
        bboxes2 (Tensor): shape (B, n, 4) in <x1, y1, x2, y2> format or empty.
            B indicates the batch dim, in shape (B1, B2, ..., Bn).
            If ``is_aligned `` is ``True``, then m and n must be equal.
        mode (str): "iou" (intersection over union) or "iof" (intersection over
            foreground).
        is_aligned (bool, optional): If True, then m and n must be equal.
            Default False.
        eps (float, optional): A value added to the denominator for numerical
            stability. Default 1e-6.
    Returns:
        Tensor: shape (m, n) if ``is_aligned `` is False else shape (m,)
    """
    assert mode in ['iou', 'iof', 'giou', 'diou'], 'Unsupported mode {}'.format(
        mode)
    # Either the boxes are empty or the length of boxes's last dimenstion is 4
    assert (bboxes1.shape[-1] == 4 or bboxes1.shape[0] == 0)
    assert (bboxes2.shape[-1] == 4 or bboxes2.shape[0] == 0)

    # Batch dim must be the same
    # Batch dim: (B1, B2, ... Bn)
    assert bboxes1.shape[:-2] == bboxes2.shape[:-2]
    batch_shape = bboxes1.shape[:-2]

    rows = bboxes1.shape[-2] if bboxes1.shape[0] > 0 else 0
    cols = bboxes2.shape[-2] if bboxes2.shape[0] > 0 else 0
    if is_aligned:
        assert rows == cols

    if rows * cols == 0:
        if is_aligned:
            return np.random.random(batch_shape + (rows, ))
        else:
            return np.random.random(batch_shape + (rows, cols))

    area1 = (bboxes1[..., 2] - bboxes1[..., 0]) * (
        bboxes1[..., 3] - bboxes1[..., 1])
    area2 = (bboxes2[..., 2] - bboxes2[..., 0]) * (
        bboxes2[..., 3] - bboxes2[..., 1])

    if is_aligned:
        lt = np.maximum(bboxes1[..., :2], bboxes2[..., :2])  # [B, rows, 2]
        rb = np.minimum(bboxes1[..., 2:], bboxes2[..., 2:])  # [B, rows, 2]

        wh = (rb - lt).clip(min=0)  # [B, rows, 2]
        overlap = wh[..., 0] * wh[..., 1]

        if mode in ['iou', 'giou']:
            union = area1 + area2 - overlap
        else:
            union = area1
        if mode == 'giou':
            enclosed_lt = np.minimum(bboxes1[..., :2], bboxes2[..., :2])
            enclosed_rb = np.maximum(bboxes1[..., 2:], bboxes2[..., 2:])
        if mode == 'diou':
            enclosed_lt = np.minimum(bboxes1[..., :2], bboxes2[..., :2])
            enclosed_rb = np.maximum(bboxes1[..., 2:], bboxes2[..., 2:])
            b1_x1, b1_y1 = bboxes1[..., 0], bboxes1[..., 1]
            b1_x2, b1_y2 = bboxes1[..., 2], bboxes1[..., 3]
            b2_x1, b2_y1 = bboxes2[..., 0], bboxes2[..., 1]
            b2_x2, b2_y2 = bboxes2[..., 2], bboxes2[..., 3]
    else:
        lt = np.maximum(bboxes1[..., :, None, :2],
                        bboxes2[..., None, :, :2])  # [B, rows, cols, 2]
        rb = np.minimum(bboxes1[..., :, None, 2:],
                        bboxes2[..., None, :, 2:])  # [B, rows, cols, 2]

        wh = (rb - lt).clip(min=0)  # [B, rows, cols, 2]
        overlap = wh[..., 0] * wh[..., 1]

        if mode in ['iou', 'giou']:
            union = area1[..., None] + area2[..., None, :] - overlap
        else:
            union = area1[..., None]
        if mode == 'giou':
            enclosed_lt = np.minimum(bboxes1[..., :, None, :2],
                                     bboxes2[..., None, :, :2])
            enclosed_rb = np.maximum(bboxes1[..., :, None, 2:],
                                     bboxes2[..., None, :, 2:])
        if mode == 'diou':
            enclosed_lt = np.minimum(bboxes1[..., :, None, :2],
                                     bboxes2[..., None, :, :2])
            enclosed_rb = np.maximum(bboxes1[..., :, None, 2:],
                                     bboxes2[..., None, :, 2:])
            b1_x1, b1_y1 = bboxes1[..., :, None, 0], bboxes1[..., :, None, 1]
            b1_x2, b1_y2 = bboxes1[..., :, None, 2], bboxes1[..., :, None, 3]
            b2_x1, b2_y1 = bboxes2[..., None, :, 0], bboxes2[..., None, :, 1]
            b2_x2, b2_y2 = bboxes2[..., None, :, 2], bboxes2[..., None, :, 3]

    eps = np.array([eps])
    union = np.maximum(union, eps)
    ious = overlap / union
    if mode in ['iou', 'iof']:
        return ious
    # calculate gious
    if mode in ['giou']:
        enclose_wh = (enclosed_rb - enclosed_lt).clip(min=0)
        enclose_area = enclose_wh[..., 0] * enclose_wh[..., 1]
        enclose_area = np.maximum(enclose_area, eps)
        gious = ious - (enclose_area - union) / enclose_area
        return gious
    if mode in ['diou']:
        left = ((b2_x1 + b2_x2) - (b1_x1 + b1_x2))**2 / 4
        right = ((b2_y1 + b2_y2) - (b1_y1 + b1_y2))**2 / 4
        rho2 = left + right
        enclose_wh = (enclosed_rb - enclosed_lt).clip(min=0)
        enclose_c = enclose_wh[..., 0]**2 + enclose_wh[..., 1]**2
        enclose_c = np.maximum(enclose_c, eps)
        dious = ious - rho2 / enclose_c
        return dious