app.py

import gradio as gr
import torch
from pathlib import Path
from PIL import Image
import pandas as pd
from lavis.models import load_model_and_preprocess
from lavis.processors import load_processor
from transformers import CLIPTokenizerFast, CLIPModel  # Import CLIPTokenizerFast

# Load model and preprocessors for Image-Text Matching (LAVIS)
device = torch.device("cuda") if torch.cuda.is_available() else "cpu"
model_itm, vis_processors, text_processors = load_model_and_preprocess("blip2_image_text_matching", "pretrain", device=device, is_eval=True)

# Load model and processor for Image Captioning (TextCaps)
model_caption = CLIPModel.from_pretrained("microsoft/git-large-r-textcaps")
processor_caption = CLIPTokenizerFast.from_pretrained("microsoft/git-large-r-textcaps", from_slow=True)  # Convert tokenizer

# List of statements for Image-Text Matching
statements = [
    "cartoon, figurine, or toy",
    "appears to be for children",
    "includes children",
    "is sexual",
    "depicts a child or portrays objects, images, or cartoon figures that primarily appeal to persons below the legal purchase age",
    "uses the name of or depicts Santa Claus",
    'promotes alcohol use as a "rite of passage" to adulthood',
]

txts = [text_processors["eval"](statement) for statement in statements]

# Function to compute Image-Text Matching (ITM) scores for all statements
def compute_itm_scores(image):
    pil_image = Image.fromarray(image.astype('uint8'), 'RGB')
    img = vis_processors["eval"](pil_image.convert("RGB")).unsqueeze(0).to(device)
    results = []
    for i, statement in enumerate(statements):
        txt = txts[i]
        itm_output = model_itm({"image": img, "text_input": txt}, match_head="itm")
        itm_scores = torch.nn.functional.softmax(itm_output, dim=1)
        score = itm_scores[:, 1].item()
        result_text = f'The image and "{statement}" are matched with a probability of {score:.3%}'
        results.append(result_text)
    output = "\n".join(results)
    return output

# Function to generate image captions using TextCaps
def generate_image_captions(image):
    pil_image = Image.fromarray(image.astype('uint8'), 'RGB')
    inputs = processor_caption(pil_image, return_tensors="pt", padding=True, truncation=True)
    outputs = model_caption.generate(**inputs)
    caption = processor_caption.decode(outputs[0])
    return caption

# Main function to perform image captioning and image-text matching
def process_images_and_statements(image):
    # Generate image captions using TextCaps
    captions = generate_image_captions(image)
    
    # Compute ITM scores for predefined statements using LAVIS
    itm_scores = compute_itm_scores(image)
    
    # Combine image captions and ITM scores into the output
    output = "Image Captions:\n" + captions + "\n\nITM Scores:\n" + itm_scores
    return output

image_input = gr.inputs.Image()
output = gr.outputs.Textbox(label="Results")

iface = gr.Interface(fn=process_images_and_statements, inputs=image_input, outputs=output, title="Image Captioning and Image-Text Matching")
iface.launch()