app.py

import os
from huggingface_hub import login
from transformers import BlipProcessor, BlipForConditionalGeneration
from PIL import Image

import gradio as gr
from diffusers import DiffusionPipeline
import torch
import spaces  # Hugging Face Spaces module

import requests
from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info









# Get Hugging Face Token from environment variable
hf_token = os.getenv('HF_AUTH_TOKEN')
if not hf_token:
    raise ValueError("Hugging Face token is not set in the environment variables.")
login(token=hf_token)



# Load the processor and model
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
processor1 = BlipProcessor.from_pretrained("noamrot/FuseCap")
model2 = BlipForConditionalGeneration.from_pretrained("noamrot/FuseCap")
pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-3.5-medium")
model3 =model = Qwen2VLForConditionalGeneration.from_pretrained(
    "prithivMLmods/Qwen2-VL-OCR-2B-Instruct", torch_dtype="auto", device_map="auto"
)
processor2 = AutoProcessor.from_pretrained("prithivMLmods/Qwen2-VL-OCR-2B-Instruct")




device = "cuda" if torch.cuda.is_available() else "cpu"
pipe.to(device)
model2.to(device)
model.to(device)



@spaces.GPU(duration=150)
def generate_caption_and_image(image):
    img = image.convert("RGB")
    # reader = easyocr.Reader(['en'])  
    # result = reader.readtext(img)
    import random
    messages = [{"role": "user", "content": [{"type": "image", "image": img}, {"type": "text", "text": "Describe this Image"}]}]

    text = processor.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )
    image_inputs, video_inputs = process_vision_info(messages)
    inputs = processor(
        text=[text],
        images=image_inputs,
        videos=video_inputs,
        padding=True,
        return_tensors="pt",
    )
    inputs = inputs.to("cuda")
    generated_ids = model.generate(**inputs, max_new_tokens=128)
    generated_ids_trimmed = [
        out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
    ]
    result = processor.batch_decode(
        generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
    )
    


        

    # Define lists for the three variables
    fabrics = ['cotton', 'silk', 'denim', 'linen', 'polyester', 'wool', 'velvet']
    patterns = ['striped', 'floral', 'geometric', 'abstract', 'solid', 'polka dots']
    textile_designs = ['woven texture', 'embroidery', 'printed fabric', 'hand-dyed', 'quilting']
    
    # Randomly select one from each category
    selected_fabric = random.choice(fabrics)
    selected_pattern = random.choice(patterns)
    selected_textile_design = random.choice(textile_designs)
    text = "a picture of "
    inputs = processor(img, text, return_tensors="pt").to(device)
    
    out = model.generate(**inputs, num_beams = 3)
   
    

    caption2 = processor.decode(out[0], skip_special_tokens=True)
    
    # Generate caption
    inputs = processor(image, return_tensors="pt", padding=True, truncation=True, max_length=250)
    inputs = {key: val.to(device) for key, val in inputs.items()}
    out = model.generate(**inputs)
    caption1 = processor.decode(out[0], skip_special_tokens=True)

    prompt = f'''Create a highly realistic clothing item based on the following descriptions: The design should reflect {caption1} and {caption2}, blending both themes into a single, stylish, and modern piece of clothing. Incorporate highly realistic and high-quality textures that exude sophistication, with realistic fabric lighting and fine details. Subtly hint at {selected_fabric}, featuring a {selected_pattern} motif and a {selected_textile_design} style that harmoniously balances the essence of both captions.and {result} should be written on top of it'''



    # Generate image based on the caption
    generated_image = pipe(prompt).images[0]

    return prompt, generated_image

# Gradio UI
iface = gr.Interface(
    fn=generate_caption_and_image,
    inputs=gr.Image(type="pil", label="Upload Image"),
    outputs=[gr.Textbox(label="Generated Caption"), gr.Image(label="Generated Design")],
    live=True
)
iface.launch(share=True)