app.py

import torch 
import re 
import gradio as gr
from transformers import AutoTokenizer, ViTFeatureExtractor, VisionEncoderDecoderModel 

device='cpu'
encoder_checkpoint = "Thibalte/captionning_project"
feature_extractor = ViTFeatureExtractor.from_pretrained(encoder_checkpoint)
tokenizer = AutoTokenizer.from_pretrained("Thibalte/captionning_project")
model = AutoModel.from_pretrained("Thibalte/captionning_project")


def predict(image,max_length=24, num_beams=4):
  image = image.convert('RGB')
  image = feature_extractor(image, return_tensors="pt").pixel_values.to(device)
  clean_text = lambda x: x.replace('<|endoftext|>','').split('\n')[0]
  caption_ids = model.generate(image, max_length = max_length)[0]
  caption_text = clean_text(tokenizer.decode(caption_ids))
  return caption_text 


# Gradio Interface
gradio_app = gr.Interface(
    fn=predict,
    inputs=gr.Image(label="Select image for captioning", sources=['upload', 'webcam'], type="pil"),
    outputs=[gr.Textbox(label="Image Caption")],
    examples = [f"example{i}.jpg" for i in range(1,7)],
    title="Image Captioning with our model",
)

if __name__ == "__main__":
    gradio_app.launch()