app.py

import gradio as gr
import spaces
import torch
from transformers import AutoProcessor, Florence2ForConditionalGeneration

MODEL_ID = "florence-community/Florence-2-large"

model = Florence2ForConditionalGeneration.from_pretrained(
    MODEL_ID, dtype=torch.float16
).to("cuda")
processor = AutoProcessor.from_pretrained(MODEL_ID)


@spaces.GPU
def read_text(image):
    if image is None:
        return "Please upload an image first."
    image = image.convert("RGB")
    inputs = processor(text="<OCR>", images=image, return_tensors="pt").to(
        "cuda", torch.float16
    )
    generated_ids = model.generate(
        **inputs, max_new_tokens=1024, num_beams=3, do_sample=False
    )
    raw = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
    parsed = processor.post_process_generation(
        raw, task="<OCR>", image_size=image.size
    )
    text = parsed["<OCR>"].strip()
    return text or "No text was found in this image."


demo = gr.Interface(
    fn=read_text,
    inputs=gr.Image(type="pil", label="Upload an image with printed text"),
    outputs=gr.Textbox(label="Text the model read", lines=10),
    title="OCR Reader",
    description="Upload an image of printed text and get back the text the model reads.",
)

demo.launch()