app.py

import tempfile
import gradio as gr
from gtts import gTTS
import inference_script
import vit_gpt2
import os
import warnings
warnings.filterwarnings('ignore')


def process_image_and_generate_output(image, model_selection):

    if model_selection == ('Basic Model (Trained only for 15 epochs without any hyperparameter tuning, utilizing '
                           'inception v3)'):
        result = inference_script.evaluate(image)
        pred_caption = ' '.join(result).rsplit(' ', 1)[0]
        pred_caption = pred_caption.replace('<unk>', '')
    elif model_selection == 'ViT-GPT2 (SOTA model for Image captioning)':
        result = vit_gpt2.predict_step(image)
        pred_caption = result[0]
    else:
        return "Invalid model selection", None

    # Generate speech from the caption
    tts = gTTS(text=pred_caption, lang='en', slow=False)
    with tempfile.NamedTemporaryFile(delete=False, suffix='.mp3') as temp_audio:
        audio_file_path = temp_audio.name
        tts.save(audio_file_path)

    # Read the audio file
    with open(audio_file_path, "rb") as f:
        audio_content = f.read()

    # Clean up the temporary audio file
    os.unlink(audio_file_path)
    return pred_caption, audio_content


iface = gr.Interface(fn=process_image_and_generate_output,
                     inputs=["image", gr.Radio(["Basic Model (Trained only for 15 epochs without any hyperparameter "
                                                "tuning, utilizing inception v3)", "ViT-GPT2 (SOTA model for Image "
                                                                                   "captioning)"], label="Choose "
                                                                                                         "Model")],
                     outputs=["text", "audio"],
                     title="Eye For Blind | Image Captioning & TTS",
                     description="To be added")

iface.launch()