图像扩展 同比例缩小，补成正方形，补黑

/mnt/pfs/users/lbg/project/4dhuman/multi-hmr/utils/vitdet_dataset.py

def rotate_2d(pt_2d: np.array, rot_rad: float) -> np.array:
    """
    Rotate a 2D point on the x-y plane.
    Args:
        pt_2d (np.array): Input 2D point with shape (2,).
        rot_rad (float): Rotation angle
    Returns:
        np.array: Rotated 2D point.
    """
    x = pt_2d[0]
    y = pt_2d[1]
    sn, cs = np.sin(rot_rad), np.cos(rot_rad)
    xx = x * cs - y * sn
    yy = x * sn + y * cs
    return np.array([xx, yy], dtype=np.float32)

def gen_trans_from_patch_cv(c_x: float, c_y: float,
                            src_width: float, src_height: float,
                            dst_width: float, dst_height: float,
                            scale: float, rot: float) -> np.array:
    """
    Create transformation matrix for the bounding box crop.
    Args:
        c_x (float): Bounding box center x coordinate in the original image.
        c_y (float): Bounding box center y coordinate in the original image.
        src_width (float): Bounding box width.
        src_height (float): Bounding box height.
        dst_width (float): Output box width.
        dst_height (float): Output box height.
        scale (float): Rescaling factor for the bounding box (augmentation).
        rot (float): Random rotation applied to the box.
    Returns:
        trans (np.array): Target geometric transformation.
    """
    # augment size with scale
    src_w = src_width * scale
    src_h = src_height * scale
    src_center = np.zeros(2)
    src_center[0] = c_x
    src_center[1] = c_y
    # augment rotation
    rot_rad = np.pi * rot / 180
    src_downdir = rotate_2d(np.array([0, src_h * 0.5], dtype=np.float32), rot_rad)
    src_rightdir = rotate_2d(np.array([src_w * 0.5, 0], dtype=np.float32), rot_rad)

    dst_w = dst_width
    dst_h = dst_height
    dst_center = np.array([dst_w * 0.5, dst_h * 0.5], dtype=np.float32)
    dst_downdir = np.array([0, dst_h * 0.5], dtype=np.float32)
    dst_rightdir = np.array([dst_w * 0.5, 0], dtype=np.float32)

    src = np.zeros((3, 2), dtype=np.float32)
    src[0, :] = src_center
    src[1, :] = src_center + src_downdir
    src[2, :] = src_center + src_rightdir

    dst = np.zeros((3, 2), dtype=np.float32)
    dst[0, :] = dst_center
    dst[1, :] = dst_center + dst_downdir
    dst[2, :] = dst_center + dst_rightdir

    trans = cv2.getAffineTransform(np.float32(src), np.float32(dst))

    return trans


def generate_image_patch_cv2(img: np.array, c_x: float, c_y: float,
                             bb_width: float, bb_height: float,
                             patch_width: float, patch_height: float,
                             do_flip: bool, scale: float, rot: float,
                             border_mode=cv2.BORDER_CONSTANT, border_value=0) -> Tuple[np.array, np.array]:
    """
    Crop the input image and return the crop and the corresponding transformation matrix.
    Args:
        img (np.array): Input image of shape (H, W, 3)
        c_x (float): Bounding box center x coordinate in the original image.
        c_y (float): Bounding box center y coordinate in the original image.
        bb_width (float): Bounding box width.
        bb_height (float): Bounding box height.
        patch_width (float): Output box width.
        patch_height (float): Output box height.
        do_flip (bool): Whether to flip image or not.
        scale (float): Rescaling factor for the bounding box (augmentation).
        rot (float): Random rotation applied to the box.
    Returns:
        img_patch (np.array): Cropped image patch of shape (patch_height, patch_height, 3)
        trans (np.array): Transformation matrix.
    """

    img_height, img_width, img_channels = img.shape
    if do_flip:
        img = img[:, ::-1, :]
        c_x = img_width - c_x - 1


    trans = gen_trans_from_patch_cv(c_x, c_y, bb_width, bb_height, patch_width, patch_height, scale, rot)

    img_patch = cv2.warpAffine(img, trans, (int(patch_width), int(patch_height)), 
                        flags=cv2.INTER_LINEAR, 
                        borderMode=border_mode,
                        borderValue=border_value,
                )
    
    cv2.imwrite("aaa.jpg",img_patch)
    # Force borderValue=cv2.BORDER_CONSTANT for alpha channel
    if (img.shape[2] == 4) and (border_mode != cv2.BORDER_CONSTANT):
        img_patch[:,:,3] = cv2.warpAffine(img[:,:,3], trans, (int(patch_width), int(patch_height)), 
                                            flags=cv2.INTER_LINEAR, 
                                            borderMode=cv2.BORDER_CONSTANT,
                            )

    return img_patch, trans

原文地址：https://blog.csdn.net/jacke121/article/details/145216151

免责声明：本站文章内容转载自网络资源，如本站内容侵犯了原著者的合法权益，可联系本站删除。更多内容请关注自学内容网（zxcms.com）！