costum_datasets.py

from configs import CFG
import os
import requests
import zipfile
from pycocotools.coco import COCO
import torch
import cv2
import albumentations as A
import soundfile as sf



# Load Coco dataset
def download_dataset(data_dir="../datasets"):
    # Create caption and image directories
    annotations_dir = os.path.join(data_dir, "annotations")
    images_dir = os.path.join(data_dir, "train2014")

    # Download annotations (captions)
    zip_file = os.path.join(annotations_dir, "annotations.zip")
    url = "http://images.cocodataset.org/annotations/annotations_trainval2014.zip"
    response = requests.get(url, stream=True)
    # write chunk in zip file
    with open(zip_file, "wb") as f:
        # 8192 = 8KB chunks (block or piece of data)
        for chunk in response.iter_content(chunk_size=8192):
            f.write(chunk)
    # unzip file
    with zipfile.ZipFile(zip_file, "r") as zip_ref:
        zip_ref.extractall(data_dir)  # Extract all contents to the specified directory
    os.remove(zip_file)

    # Download train images
    zip_file = os.path.join(images_dir, "train2014.zip")
    url = "http://images.cocodataset.org/zips/train2014.zip"
    response = requests.get(url, stream=True)
    # write chunk in zip file
    with open(zip_file, "wb") as f:
        for chunk in response.iter_content(chunk_size=8192):
            f.write(chunk)
    # unzip file
    with zipfile.ZipFile(zip_file, "r") as zip_ref:
        zip_ref.extractall(data_dir)  # Extract all contents to the specified directory
    os.remove(zip_file)

    # Download val images
    images_dir = os.path.join(data_dir, "val2014")
    zip_file = os.path.join(images_dir, "val2014.zip")
    url = "http://images.cocodataset.org/zips/val2014.zip"
    response = requests.get(url, stream=True)
    # write chunk in zip file
    with open(zip_file, "wb") as f:
        for chunk in response.iter_content(chunk_size=8192):
            f.write(chunk)
    # unzip file
    with zipfile.ZipFile(zip_file, "r") as zip_ref:
        zip_ref.extractall(data_dir)  # Extract all contents to the specified directory
    os.remove(zip_file)



def make_pairs(annotation_json_files, image_dir, max_captions=3):
    
    images = os.listdir(annotation_json_files)
    image_caption = [(os.path.join(annotation_json_files, image), "an image") for image in images]
    return image_caption