app.py

import gradio as gr
from PIL import Image
import torch
from transformers import AutoProcessor, AutoModelForCausalLM

# Load Florence-2 (runs on CPU, free tier compatible)
model_id = "microsoft/Florence-2-large"
device = "cuda" if torch.cuda.is_available() else "cpu"
dtype = torch.float16 if torch.cuda.is_available() else torch.float32

print(f"Loading model on {device}...")
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=dtype,
    trust_remote_code=True
).to(device)
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
print("Model loaded.")

def analyze_image(image, prompt):
    if image is None:
        return "No image uploaded."
    
    if not prompt:
        prompt = "<MORE_DETAILED_CAPTION>"
    
    inputs = processor(text=prompt, images=image, return_tensors="pt").to(device, dtype)
    
    with torch.no_grad():
        generated_ids = model.generate(
            input_ids=inputs["input_ids"],
            pixel_values=inputs["pixel_values"],
            max_new_tokens=512,
            do_sample=False
        )
    
    generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
    parsed = processor.post_process_generation(generated_text, task=prompt, image_size=(image.width, image.height))
    
    # Return the first value from parsed dict
    if isinstance(parsed, dict):
        return list(parsed.values())[0]
    return str(parsed)

# Available tasks for Florence-2
TASKS = [
    "<CAPTION>",
    "<DETAILED_CAPTION>",
    "<MORE_DETAILED_CAPTION>",
    "<OCR>",
    "<OCR_WITH_REGION>",
    "<OBJECT_DETECTION>",
    "<REGION_TO_CATEGORY>",
    "<REGION_TO_DESCRIPTION>",
]

with gr.Blocks(title="Vision Analyzer") as demo:
    gr.Markdown("# Image Understanding")
    gr.Markdown("Upload an image and select what you want to extract from it.")
    
    with gr.Row():
        with gr.Column():
            image_input = gr.Image(type="pil", label="Upload Image")
            task_dropdown = gr.Dropdown(choices=TASKS, value="<MORE_DETAILED_CAPTION>", label="Analysis Type")
            text_prompt = gr.Textbox(label="Or enter custom prompt (overrides dropdown)", placeholder="Describe what you see...", lines=2)
            analyze_btn = gr.Button("Analyze")
        with gr.Column():
            output = gr.Textbox(label="Result", lines=15, show_copy_button=True)
    
    analyze_btn.click(fn=analyze_image, inputs=[image_input, text_prompt], outputs=output)
    
    gr.Markdown("---")
    gr.Markdown("Powered by Microsoft Florence-2-large on HuggingFace free tier.")

demo.launch(server_name="0.0.0.0", server_port=7860)