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 image is None:
        return "Please select an image", None

    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


# Define your sample images
# sample_images = [os.path.join(os.path.dirname(__file__), 'sample_images/1.jpg'),
#                  os.path.join(os.path.dirname(__file__), 'sample_images/2.jpg'),
#                  os.path.join(os.path.dirname(__file__), 'sample_images/3.jpg'),
#                  os.path.join(os.path.dirname(__file__), 'sample_images/4.jpg'), ]
sample_images = [
    ["sample_images/1.jpg"],
    ["sample_images/2.jpg"],
    ["sample_images/3.jpg"],
    ["sample_images/4.jpg"]
]

# Create a dropdown to select sample image
image_input = gr.Image(label="Upload Image", sources=['upload', 'webcam'])

# Create a dropdown to choose the model
model_selection_input = 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")

iface = gr.Interface(fn=process_image_and_generate_output,
                     inputs=[image_input, model_selection_input],
                     outputs=["text", "audio"],
                     examples=sample_images,
                     title="Eye For Blind | Image Captioning & TTS",
                     description="To be added")

iface.launch()