from transformers import pipeline
from transformers.utils import logging
from helper import ignore_warnings, load_image_from_url, render_results_in_image, summarize_predictions_natural_language

import gradio as gr

logging.set_verbosity_error()
ignore_warnings()

od_pipe = pipeline("object-detection", model="facebook/detr-resnet-50")
tts_pipe = pipeline("text-to-speech", model="kakao-enterprise/vits-ljs")

def detect_objects(pil_image):
    pipeline_output = od_pipe(pil_image)
    processed_image = render_results_in_image(pil_image, pipeline_output)
    text = summarize_predictions_natural_language(pipeline_output)
    narrated_text = tts_pipe(text)
    sr=narrated_text["sampling_rate"]
    audio = narrated_text["audio"][0]
    
    return processed_image, text,(sr, audio)

demo = gr.Interface(title="Object Detection in an Image and Narration - test & demo app by Srinivas.V..",
                    description="Upload any image,preferably an image with many clearly distinguishable objects and submit. Play the audio to listen",
  fn=detect_objects,
  inputs=gr.Image(label="Input image",
                  type="pil"),
  outputs=[gr.Image(label="Output image with predicted instances",
                   type="pil"), gr.Textbox(label='Description of detected objects', lines=3),
            gr.Audio(label='Play audio to listen about the detected objectes in the image')]
)

demo.launch(debug=True, share=True)