آماده‌سازی دیتاست LoRA با Jupyter Notebook

mkdir lora_dataset

lora_dataset/
  image1.jpg
  image2.jpg
  image3.jpg
  image4.jpg

!pip install pillow tqdm opencv-python

from PIL import Image
import os

input_folder = "lora_dataset"
output_folder = "dataset_resized"

os.makedirs(output_folder, exist_ok=True)

size = (1024, 1024)

for file in os.listdir(input_folder):
    if file.endswith((".png",".jpg",".jpeg")):
        img = Image.open(os.path.join(input_folder, file)).convert("RGB")
        img = img.resize(size)
        img.save(os.path.join(output_folder, file))

def center_crop(img):
    w, h = img.size
    min_dim = min(w, h)

    left = (w - min_dim) // 2
    top = (h - min_dim) // 2
    right = left + min_dim
    bottom = top + min_dim

    return img.crop((left, top, right, bottom))

img = Image.open("example.jpg")
img = center_crop(img)
img = img.resize((1024,1024))

image1.jpg
image1.txt

portrait of a cyberpunk female character, neon lighting

import os

dataset_path = "dataset_resized"

base_caption = "photo of sks person"

for file in os.listdir(dataset_path):
    if file.endswith(".jpg") or file.endswith(".png"):
        name = os.path.splitext(file)[0]
        caption_file = os.path.join(dataset_path, name + ".txt")

        with open(caption_file,"w") as f:
            f.write(base_caption)

!pip install transformers

from transformers import BlipProcessor, BlipForConditionalGeneration
from PIL import Image

processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")

img = Image.open("example.jpg").convert("RGB")

inputs = processor(img, return_tensors="pt")
out = model.generate(**inputs)

caption = processor.decode(out[0], skip_special_tokens=True)

print(caption)

dataset/
  image1.jpg
  image1.txt
  image2.jpg
  image2.txt
  image3.jpg
  image3.txt