import torch 
import re
from PIL import Image
import requests
import streamlit as st
from transformers import AutoTokenizer, ViTFeatureExtractor, VisionEncoderDecoderModel 


st.set_page_config(page_title="Captionize")

st.title("🤖 Captionize")
st.subheader("Generate Captions for your Image...")

st.sidebar.image('./csv_analysis.png',width=300, use_column_width=True)

# Applying Styling
st.markdown("""
<style>
div.stButton > button:first-child {
    background-color: #0099ff;
    color:#ffffff;
}
div.stButton > button:hover {
    background-color: #00ff00;
    color:#FFFFFF;
    }
</style>""", unsafe_allow_html=True)


device='cpu'
encoder_checkpoint = "nlpconnect/vit-gpt2-image-captioning"
decoder_checkpoint = "nlpconnect/vit-gpt2-image-captioning"
model_checkpoint = "nlpconnect/vit-gpt2-image-captioning"
feature_extractor = ViTFeatureExtractor.from_pretrained(encoder_checkpoint)
tokenizer = AutoTokenizer.from_pretrained(decoder_checkpoint)
model = VisionEncoderDecoderModel.from_pretrained(model_checkpoint).to(device)

def predict(image,max_length=64, num_beams=4):
  #image = image.convert('RGB')
  image = Image.open(requests.get(image, stream=True).raw).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

pic = st.file_uploader(label="Please upload any Image here 😎",type=['png', 'jpeg', 'jpg'], help="Only 'png', 'jpeg' or 'jpg' formats allowed")


button = st.button("Generate Caption")

if button:
    # Get Response
    caption =  predict(pic)
    st.write(caption)