image_processing_ernie_45t_vl.py

# Copyright (c) 2025 Baidu, Inc. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

"""Image processor class for Ernie_45T_VL."""

import math
from typing import List, Optional, Union
from PIL import Image
import numpy as np

from transformers.image_processing_utils import BaseImageProcessor, BatchFeature
from transformers.image_transforms import (
    convert_to_rgb,
    normalize,
    rescale,
    resize,
    to_channel_dimension_format,
)
from transformers.image_utils import (
    OPENAI_CLIP_MEAN,
    OPENAI_CLIP_STD,
    ChannelDimension,
    ImageInput,
    PILImageResampling,
    get_image_size,
    infer_channel_dimension_format,
    is_valid_image,
    make_list_of_images,
    to_numpy_array,
    valid_images,
)
from transformers.utils import TensorType, logging
from transformers.video_utils import VideoInput


logger = logging.get_logger(__name__)


def round_by_factor(number: int, factor: int) -> int:
    """Returns the closest integer to 'number' that is divisible by 'factor'."""
    return round(number / factor) * factor


def ceil_by_factor(number: int, factor: int) -> int:
    """Returns the smallest integer greater than or equal to 'number' that is divisible by 'factor'."""
    return math.ceil(number / factor) * factor


def floor_by_factor(number: int, factor: int) -> int:
    """Returns the largest integer less than or equal to 'number' that is divisible by 'factor'."""
    return math.floor(number / factor) * factor


def smart_resize(
    height: int,
    width: int,
    factor: int = 28,
    min_pixels: int = 4 * 28 * 28,
    max_pixels: int = 16384 * 28 * 28,
):
    """
    Rescales the image so that the following conditions are met:

    1. Both dimensions (height and width) are divisible by 'factor'.

    2. The total number of pixels is within the range ['min_pixels', 'max_pixels'].

    3. The aspect ratio of the image is maintained as closely as possible.
    """
    MAX_RATIO = 200
    if max(height, width) / min(height, width) > MAX_RATIO:
        if height > width:
            new_width = max(factor, round_by_factor(width, factor))
            new_height = floor_by_factor(new_width * MAX_RATIO, factor)
        else:
            new_height = max(factor, round_by_factor(height, factor))
            new_width = floor_by_factor(new_height * MAX_RATIO, factor)

        logger.info(
            f"absolute aspect ratio must be smaller than {MAX_RATIO}, got {max(height, width) / min(height, width)},\
              resize to {max(new_height, new_width) / min(new_height, new_width)}"
        )

        height = new_height
        width = new_width

    h_bar = max(factor, round_by_factor(height, factor))
    w_bar = max(factor, round_by_factor(width, factor))
    if h_bar * w_bar > max_pixels:
        beta = math.sqrt((height * width) / max_pixels)
        h_bar = floor_by_factor(height / beta, factor)
        w_bar = floor_by_factor(width / beta, factor)
    elif h_bar * w_bar < min_pixels:
        beta = math.sqrt(min_pixels / (height * width))
        h_bar = ceil_by_factor(height * beta, factor)
        w_bar = ceil_by_factor(width * beta, factor)

    if min_pixels > h_bar * w_bar or h_bar * w_bar > max_pixels:
        raise ValueError(f"encounter invalid h_bar: {h_bar}, w_bar: {w_bar}")

    return h_bar, w_bar


def is_scaled_image(image: np.ndarray) -> bool:
    """
    Checks to see whether the pixel values have already been rescaled to [0, 1].
    """
    if image.dtype == np.uint8:
        return False

    # It's possible the image has pixel values in [0, 255] but is of floating type
    return np.min(image) >= 0 and np.max(image) <= 1


def make_batched_images(images) -> List[List[ImageInput]]:
    """
    Accepts images in list or nested list format, and makes a list of images for preprocessing.

    Args:
        images (`Union[List[List[ImageInput]], List[ImageInput], ImageInput]`):
            The input image.

    Returns:
        list: A list of images.
    """
    if (
        isinstance(images, (list, tuple))
        and isinstance(images[0], (list, tuple))
        and is_valid_image(images[0][0])
    ):
        return [img for img_list in images for img in img_list]

    elif isinstance(images, (list, tuple)) and is_valid_image(images[0]):
        return images

    elif is_valid_image(images):
        return [images]

    raise ValueError(f"Could not make batched images from {images}")


# Copied from transformers.models.llava_next_video.image_processing_llava_next_video.make_batched_videos
def make_batched_videos(videos) -> List[VideoInput]:
    """dummy"""
    if (
        isinstance(videos, (list, tuple))
        and isinstance(videos[0], (list, tuple))
        and is_valid_image(videos[0][0])
    ):
        return videos

    elif isinstance(videos, (list, tuple)) and is_valid_image(videos[0]):
        if isinstance(videos[0], Image.Image):
            return [videos]
        elif len(videos[0].shape) == 4:
            return [list(video) for video in videos]

    elif is_valid_image(videos) and len(videos.shape) == 4:
        return [list(videos)]

    raise ValueError(f"Could not make batched video from {videos}")


class Ernie_45T_VLImageProcessor(BaseImageProcessor):
    r"""
    Constructs a adaptive image processor that dynamically resizes images based on the original images.

    Args:
        do_resize (`bool`, *optional*, defaults to `True`):
            Whether to resize the image's (height, width) dimensions.
        resample (`PILImageResampling`, *optional*, defaults to `Resampling.BICUBIC`):
            Resampling filter to use when resizing the image.
        do_rescale (`bool`, *optional*, defaults to `True`):
            Whether to rescale the image by the specified scale `rescale_factor`.
        rescale_factor (`int` or `float`, *optional*, defaults to `1/255`):
            Scale factor to use if rescaling the image.
        do_normalize (`bool`, *optional*, defaults to `True`):
            Whether to normalize the image.
        image_mean (`float` or `List[float]`, *optional*, defaults to `[0.48145466, 0.4578275, 0.40821073]`):
            Mean to use if normalizing the image. This is a float or list of floats for each channel in the image.
        image_std (`float` or `List[float]`, *optional*, defaults to `[0.26862954, 0.26130258, 0.27577711]`):
            Standard deviation to use if normalizing the image. This is a float or list of floats for each channel
            in the image.
        do_convert_rgb (`bool`, *optional*, defaults to `True`):
            Whether to convert the image to RGB.
        min_pixels (`int`, *optional*, defaults to `56 * 56`):
            The min pixels of the image to resize the image.
        max_pixels (`int`, *optional*, defaults to `28 * 28 * 1280`):
            The max pixels of the image to resize the image.
        patch_size (`int`, *optional*, defaults to 14):
            The spacial patch size of the vision encoder.
        temporal_conv_size (`int`, *optional*, defaults to 2):
            The temporal conv size in resampler.
        merge_size (`int`, *optional*, defaults to 2):
            The merge size of the vision encoder to llm encoder.
    """

    model_input_names = [
        "pixel_values",
        "image_grid_thw",
        "pixel_values_videos",
        "video_grid_thw",
    ]

    def __init__(
        self,
        do_resize: bool = True,
        resample: PILImageResampling = PILImageResampling.BICUBIC,
        do_rescale: bool = True,
        rescale_factor: Union[float, List[float]] = 1 / 255,
        do_normalize: bool = True,
        image_mean: Optional[Union[float, List[float]]] = None,
        image_std: Optional[Union[float, List[float]]] = None,
        do_convert_rgb: bool = True,
        min_pixels: int = 56 * 56,
        max_pixels: int = 28 * 28 * 1280,
        patch_size: int = 14,
        temporal_conv_size: int = 2,
        merge_size: int = 2,
        **kwargs,
    ) -> None:
        """init"""
        super().__init__(**kwargs)
        self.do_resize = do_resize
        self.resample = resample
        self.do_rescale = do_rescale
        self.rescale_factor = rescale_factor
        self.do_normalize = do_normalize
        self.image_mean = image_mean if image_mean is not None else OPENAI_CLIP_MEAN
        self.image_std = image_std if image_std is not None else OPENAI_CLIP_STD
        self.min_pixels = min_pixels
        self.max_pixels = max_pixels
        self.patch_size = patch_size
        self.temporal_conv_size = temporal_conv_size
        self.merge_size = merge_size
        self.size = {"min_pixels": min_pixels, "max_pixels": max_pixels}
        self.do_convert_rgb = do_convert_rgb

    def set_pixels(self, min_pixels=None, max_pixels=None, msg=""):
        """set_pixels"""
        if min_pixels is not None:
            assert (
                isinstance(min_pixels, int) and min_pixels >= 0
            ), "min_pixels must be positive int"
            logger.info(
                f"{msg} Ernie_45T_VLImageProcessor set min_pixels = {min_pixels}"
            )
            self.min_pixels = min_pixels
            self.size["min_pixels"] = int(min_pixels)
        if max_pixels is not None:
            assert (
                isinstance(max_pixels, int) and max_pixels > 0
            ), "max_pixels must be positive int"
            logger.info(
                f"{msg} Ernie_45T_VLImageProcessor set max_pixels = {max_pixels}"
            )
            self.max_pixels = max_pixels
            self.size["max_pixels"] = int(max_pixels)

    def get_smarted_resize(self, height, width, min_pixels=None, max_pixels=None):
        """dummy"""
        actual_min_pixels = min_pixels if min_pixels is not None else self.min_pixels
        actual_max_pixels = max_pixels if max_pixels is not None else self.max_pixels
        resized_height, resized_width = smart_resize(
            height,
            width,
            factor=self.patch_size * self.merge_size,
            min_pixels=actual_min_pixels,
            max_pixels=actual_max_pixels,
        )
        return (resized_height, resized_width), (
            resized_height // self.patch_size,
            resized_width // self.patch_size,
        )

    def _preprocess(
        self,
        images: Union[ImageInput, VideoInput],
        do_resize: bool = True,
        resample: PILImageResampling = None,
        do_rescale: bool = True,
        rescale_factor: float = 1 / 255,
        do_normalize: bool = True,
        image_mean: Optional[Union[float, List[float]]] = None,
        image_std: Optional[Union[float, List[float]]] = None,
        do_convert_rgb: bool = False,
        data_format: Optional[ChannelDimension] = ChannelDimension.FIRST,
        input_data_format: Optional[Union[str, ChannelDimension]] = None,
        predetermined_grid_thw=None,
    ):
        """
        Preprocess an image or batch of images. Copy of the `preprocess` method from `CLIPImageProcessor`.

        Args:
            images (`ImageInput` or `VideoInput`):
                Image or batch of images to preprocess. Expects pixel values ranging from 0 to 255.
                If pixel values range from 0 to 1, set `do_rescale=False`.
            do_resize (`bool`, *optional*, defaults to `self.do_resize`):
                Whether to resize the image.
            resample (`PILImageResampling`, *optional*, defaults to `self.resample`):
                Resampling filter to use if resizing the image. This can be one of the `PILImageResampling` enums.
            do_rescale (`bool`, *optional*, defaults to `self.do_rescale`):
                Whether to rescale the image.
            rescale_factor (`float`, *optional*, defaults to `self.rescale_factor`):
                Scale factor to use if rescaling the image.
            do_normalize (`bool`, *optional*, defaults to `self.do_normalize`):
                Whether to normalize the image.
            image_mean (`float` or `List[float]`, *optional*, defaults to `self.image_mean`):
                Mean to use if normalizing the image.
                Can be a float or a list of floats corresponding to the number of channels in the image.
            image_std (`float` or `List[float]`, *optional*, defaults to `self.image_std`):
                Standard deviation to use if normalizing the image.
                Can be a float or a list of floats corresponding to the number of channels in the image.
            do_convert_rgb (`bool`, *optional*, defaults to `self.do_convert_rgb`):
                Whether to convert the image to RGB.
            data_format (`ChannelDimension`, *optional*, defaults to `ChannelDimension.FIRST`):
                The channel dimension format for the output image. Can be one of:
                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
                - Unset: Use the channel dimension format of the input image.
            input_data_format (`ChannelDimension` or `str`, *optional*):
                The channel dimension format for the input image. Can be one of:
                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
                - `"none"` or `ChannelDimension.NONE`: image in (height, width) format.
                - `"none"` or `ChannelDimension.NONE`: image in (height, width) format.
        """
        images = make_list_of_images(images)

        if do_convert_rgb:
            images = [convert_to_rgb(image) for image in images]

        # All transformations expect numpy arrays.
        images = [to_numpy_array(image) for image in images]

        if is_scaled_image(images[0]) and do_rescale:
            logger.warning_once(
                "It looks like you are trying to rescale already rescaled images. If the input"
                " images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again."
            )
        if input_data_format is None:
            # We assume that all images have the same channel dimension format.
            input_data_format = infer_channel_dimension_format(images[0])

        height, width = get_image_size(images[0], channel_dim=input_data_format)
        resized_height, resized_width = height, width
        processed_images = []

        if predetermined_grid_thw is not None:
            assert len(predetermined_grid_thw) == len(
                images
            ), f"len(predetermined_grid_thw) {len(predetermined_grid_thw)} == len(images) {len(images)}"

        for img_idx, image in enumerate(images):
            if do_resize:
                if predetermined_grid_thw is not None:
                    (resized_height, resized_width) = predetermined_grid_thw[img_idx]
                    resized_height *= self.patch_size
                    resized_width *= self.patch_size
                else:
                    resized_height, resized_width = smart_resize(
                        height,
                        width,
                        factor=self.patch_size * self.merge_size,
                        min_pixels=self.min_pixels,
                        max_pixels=self.max_pixels,
                    )

                image = resize(
                    image,
                    size=(resized_height, resized_width),
                    resample=resample,
                    data_format=input_data_format,
                )
            if do_rescale:
                image = rescale(
                    image, scale=rescale_factor, data_format=input_data_format
                )

            if do_normalize:
                image = normalize(
                    image=image,
                    mean=image_mean,
                    std=image_std,
                    data_format=input_data_format,
                )

            image = to_channel_dimension_format(
                image, data_format, input_channel_dim=input_data_format
            )  # [C, H, W]

            processed_images.append(image)
        patches = np.array(processed_images)
        if data_format == ChannelDimension.LAST:
            patches = patches.transpose([0, 3, 1, 2])

        channel = patches.shape[1]  # [time, C, H, W]
        grid_t = patches.shape[0]
        grid_h, grid_w = (
            resized_height // self.patch_size,
            resized_width // self.patch_size,
        )
        patches = patches.reshape(
            [
                grid_t,
                channel,
                grid_h // self.merge_size,
                self.merge_size,
                self.patch_size,
                grid_w // self.merge_size,
                self.merge_size,
                self.patch_size,
            ]
        )
        # [grid_t, grid_h/merge_size, grid_w/merge_size, merge_size, merge_size, C, psz, psz]
        patches = patches.transpose([0, 2, 5, 3, 6, 1, 4, 7])

        flatten_patches = patches.reshape(
            [grid_t * grid_h * grid_w, channel * self.patch_size * self.patch_size]
        )  # [grid_t * grid_h * grid_w, C * psz * psz]

        return flatten_patches, (grid_t, grid_h, grid_w)

    def preprocess(
        self,
        images: ImageInput,
        videos: VideoInput = None,
        do_resize: bool = True,
        size: Optional[Union[int, List[int]]] = None,
        resample: PILImageResampling = None,
        do_rescale: bool = True,
        rescale_factor: float = 1 / 255,
        do_normalize: bool = True,
        image_mean: Optional[Union[float, List[float]]] = None,
        image_std: Optional[Union[float, List[float]]] = None,
        do_convert_rgb: bool = False,
        return_tensors: Optional[Union[str, TensorType]] = None,
        data_format: Optional[ChannelDimension] = ChannelDimension.FIRST,
        input_data_format: Optional[Union[str, ChannelDimension]] = None,
        predetermined_grid_thw=None,
    ):
        """
        Args:
            images (`ImageInput`):
                Image to preprocess. Expects a single or batch of images with pixel values ranging from 0 to 255. If
                passing in images with pixel values between 0 and 1, set `do_rescale=False`.
            videos (`VideoInput`):
                Video to preprocess. Expects a single or batch of videos with pixel values ranging from 0 to 255. If
                passing in videos with pixel values between 0 and 1, set `do_rescale=False`.
            do_resize (`bool`, *optional*, defaults to `self.do_resize`):
                Whether to resize the image.
            size (`Dict[str, int]`, *optional*, defaults to `self.size`):
                Size of the image after resizing. Shortest edge of the image is resized to size["shortest_edge"], with
                the longest edge resized to keep the input aspect ratio.
            resample (`int`, *optional*, defaults to `self.resample`):
                Resampling filter to use if resizing the image. This can be one of the enum `PILImageResampling`. Only
                has an effect if `do_resize` is set to `True`.
            do_rescale (`bool`, *optional*, defaults to `self.do_rescale`):
                Whether to rescale the image.
            rescale_factor (`float`, *optional*, defaults to `self.rescale_factor`):
                Rescale factor to rescale the image by if `do_rescale` is set to `True`.
            do_normalize (`bool`, *optional*, defaults to `self.do_normalize`):
                Whether to normalize the image.
            image_mean (`float` or `List[float]`, *optional*, defaults to `self.image_mean`):
                Image mean to use for normalization. Only has an effect if `do_normalize` is set to `True`.
            image_std (`float` or `List[float]`, *optional*, defaults to `self.image_std`):
                Image standard deviation to use for normalization. Only has an effect if `do_normalize` is set to
                `True`.
            do_convert_rgb (`bool`, *optional*, defaults to `self.do_convert_rgb`):
                Whether to convert the image to RGB.
            return_tensors (`str` or `TensorType`, *optional*):
                The type of tensors to return. Can be one of:
                - Unset: Return a list of `np.ndarray`.
                - `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`.
                - `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`.
            data_format (`ChannelDimension` or `str`, *optional*, defaults to `ChannelDimension.FIRST`):
                The channel dimension format for the output image. Can be one of:
                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
                - Unset: Use the channel dimension format of the input image.
            input_data_format (`ChannelDimension` or `str`, *optional*):
                The channel dimension format for the input image. If unset, the channel dimension format is inferred
                from the input image. Can be one of:
                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
                - `"none"` or `ChannelDimension.NONE`: image in (height, width) format.

        """
        do_resize = do_resize if do_resize is not None else self.do_resize
        size = size if size is not None else self.size
        resample = resample if resample is not None else self.resample
        do_rescale = do_rescale if do_rescale is not None else self.do_rescale
        rescale_factor = (
            rescale_factor if rescale_factor is not None else self.rescale_factor
        )
        do_normalize = do_normalize if do_normalize is not None else self.do_normalize
        image_mean = image_mean if image_mean is not None else self.image_mean
        image_std = image_std if image_std is not None else self.image_std
        do_convert_rgb = (
            do_convert_rgb if do_convert_rgb is not None else self.do_convert_rgb
        )

        if images is not None:
            images = make_batched_images(images)

        if images is not None and not valid_images(images):
            raise ValueError(
                "Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, "
                "torch.Tensor."
            )

        data = {}
        if images is not None:
            pixel_values, vision_grid_thws = [], []
            for img_idx, image in enumerate(images):
                if predetermined_grid_thw is not None:
                    predetermined_grid_thw_one = [predetermined_grid_thw[img_idx]]
                else:
                    predetermined_grid_thw_one = None
                patches, image_grid_thw = self._preprocess(
                    image,
                    do_resize=do_resize,
                    resample=resample,
                    do_rescale=do_rescale,
                    rescale_factor=rescale_factor,
                    do_normalize=do_normalize,
                    image_mean=image_mean,
                    image_std=image_std,
                    data_format=data_format,
                    do_convert_rgb=do_convert_rgb,
                    input_data_format=input_data_format,
                    predetermined_grid_thw=predetermined_grid_thw_one,
                )
                pixel_values.extend(patches)
                vision_grid_thws.append(image_grid_thw)
            pixel_values = np.array(pixel_values)
            vision_grid_thws = np.array(vision_grid_thws)
            data.update(
                {"pixel_values": pixel_values, "image_grid_thw": vision_grid_thws}
            )

        if videos is not None:
            videos = make_batched_videos(videos)
            pixel_values, vision_grid_thws = [], []
            for images in videos:
                patches, video_grid_thw = self._preprocess(
                    images,
                    do_resize=do_resize,
                    resample=resample,
                    do_rescale=do_rescale,
                    rescale_factor=rescale_factor,
                    do_normalize=do_normalize,
                    image_mean=image_mean,
                    image_std=image_std,
                    data_format=data_format,
                    do_convert_rgb=do_convert_rgb,
                    input_data_format=input_data_format,
                    predetermined_grid_thw=predetermined_grid_thw,
                )
                pixel_values.extend(patches)
                vision_grid_thws.append(video_grid_thw)
            pixel_values = np.array(pixel_values)
            vision_grid_thws = np.array(vision_grid_thws)

            data.update(
                {
                    "pixel_values_videos": pixel_values,
                    "video_grid_thw": vision_grid_thws,
                }
            )

        return BatchFeature(data=data, tensor_type=return_tensors)


__all__ = ["Ernie_45T_VLImageProcessor"]