data.py

import torch
from datasets import load_dataset
from torch.utils.data import Dataset
from torchvision.transforms.functional import InterpolationMode
import torchvision.transforms as T
from logutil import get_logger


IMAGENET_MEAN = (0.485, 0.456, 0.406)
IMAGENET_STD = (0.229, 0.224, 0.225)

def build_transform(input_size):
    MEAN, STD = IMAGENET_MEAN, IMAGENET_STD
    transform = T.Compose([
        T.Lambda(lambda img: img.convert('RGB') if img.mode != 'RGB' else img),
        T.Resize((input_size, input_size), interpolation=InterpolationMode.BICUBIC),
        T.ToTensor(),
        T.Normalize(mean=MEAN, std=STD)
    ])
    return transform

def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height, image_size):
    best_ratio_diff = float('inf')
    best_ratio = (1, 1)
    area = width * height
    for ratio in target_ratios:
        target_aspect_ratio = ratio[0] / ratio[1]
        ratio_diff = abs(aspect_ratio - target_aspect_ratio)
        if ratio_diff < best_ratio_diff:
            best_ratio_diff = ratio_diff
            best_ratio = ratio
        elif ratio_diff == best_ratio_diff:
            if area > 0.5 * image_size * image_size * ratio[0] * ratio[1]:
                best_ratio = ratio
    return best_ratio

def dynamic_preprocess(image, min_num=1, max_num=12, image_size=448, use_thumbnail=False):
    orig_width, orig_height = image.size
    aspect_ratio = orig_width / orig_height

    # calculate the existing image aspect ratio
    target_ratios = set(
        (i, j) for n in range(min_num, max_num + 1) for i in range(1, n + 1) for j in range(1, n + 1) if
        i * j <= max_num and i * j >= min_num)
    target_ratios = sorted(target_ratios, key=lambda x: x[0] * x[1])

    # find the closest aspect ratio to the target
    target_aspect_ratio = find_closest_aspect_ratio(
        aspect_ratio, target_ratios, orig_width, orig_height, image_size)

    # calculate the target width and height
    target_width = image_size * target_aspect_ratio[0]
    target_height = image_size * target_aspect_ratio[1]
    blocks = target_aspect_ratio[0] * target_aspect_ratio[1]

    # resize the image
    resized_img = image.resize((target_width, target_height))
    processed_images = []
    for i in range(blocks):
        box = (
            (i % (target_width // image_size)) * image_size,
            (i // (target_width // image_size)) * image_size,
            ((i % (target_width // image_size)) + 1) * image_size,
            ((i // (target_width // image_size)) + 1) * image_size
        )
        # split the image
        split_img = resized_img.crop(box)
        processed_images.append(split_img)
    assert len(processed_images) == blocks
    if use_thumbnail and len(processed_images) != 1:
        thumbnail_img = image.resize((image_size, image_size))
        processed_images.append(thumbnail_img)
    return processed_images

def process_image(image, input_size=448, max_num=12):
    transform = build_transform(input_size=input_size)
    images = dynamic_preprocess(image, image_size=input_size, use_thumbnail=True, max_num=max_num)
    pixel_values = [transform(image) for image in images]
    pixel_values = torch.stack(pixel_values)
    return pixel_values

class BaseDataset(Dataset):
    def __init__(self, split):
        self._split = split
        self.data = []
        self.task_prompt = ""

    def __len__(self):
        return len(self.data)

    def correct_casing_finqa(self, text, is_question=False):
        if text and text[0].islower():
            text = text.capitalize()
        if not text.endswith(".") and not is_question:
            text += "."
        if not text.endswith("?") and is_question:
            text += "?"
        return text


class DocVQADataset(BaseDataset):
    """
    To have a quick look at some specific sample, e.g. a question id is 53582 and it's in validation set.
    https://huggingface.co/datasets/zhangfaen/DocumentVQA/viewer/default/validation?f[questionId][min]=53582&f[questionId][imax]=53582
    """
    def __init__(self, split):
        super().__init__(split)
        get_logger().info(f"Loading {split} dataset of zhangfaen/DocumentVQA")
        self.data = load_dataset("zhangfaen/DocumentVQA", split=split)
        get_logger().info(f"Loaded {len(self.data)} samples from {split} of zhangfaen/DocumentVQA")

    def __getitem__(self, idx):
        example = self.data[idx]
        question = self.correct_casing_finqa(example["question"], True)
        first_answer = example["answers"][0]
        answer = self.correct_casing_finqa(first_answer)
        image = example["image"]  # The image is already a PIL Image object
        if image.mode != "RGB":
            image = image.convert("RGB")
        
        assert '<image>' not in question
        question = '<image>\n' + question
        pixel_values = process_image(image)

        return {
            'question': question, 
            'answer': answer, 
            'pixel_values': pixel_values, 
            'questionId': example['questionId'], 
            'image': image
        }
    
if __name__ == "__main__":
    dataset = DocVQADataset("train")
    print(dataset[0])