app.py

import re 
import gradio as gr
from PIL import Image
from transformers import AutoProcessor, AutoModelForCausalLM
import sys,os
sys.path.append(os.getcwd())
from knowledge_extraction import get_entities, get_relations
device='cpu'

processor = AutoProcessor.from_pretrained("microsoft/git-base")
model = AutoModelForCausalLM.from_pretrained("nkasmanoff/git-planet").to(device)


def predict(image,max_length=64,device='cpu'):
    pixel_values = processor(images=image, return_tensors="pt").to(device).pixel_values
    generated_ids = model.generate(pixel_values=pixel_values, max_length=max_length)
    generated_caption = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
    relation = get_relations(generated_caption)

    entity_pair = get_entities(generated_caption)

    knowlege_triplet = f"{entity_pair[0]}-{relation}->{entity_pair[1]}"

    return knowlege_triplet 


input = gr.inputs.Image(label="Please upload a remote sensing image", type = 'pil', optional=True)
output = gr.outputs.Textbox(type="text",label="Captions")


title = "Satellite Image Captioning"

interface = gr.Interface(
        fn=predict,
        inputs = input,
        theme="grass",
        outputs=output,
        title=title,
    )
interface.launch(debug=True)