Demo.py

import streamlit as st
import sparknlp
import os
import pandas as pd

from sparknlp.base import *
from sparknlp.annotator import *
from pyspark.ml import Pipeline
from sparknlp.pretrained import PretrainedPipeline

# Page configuration
st.set_page_config(
    layout="wide", 
    initial_sidebar_state="auto"
)

# CSS for styling
st.markdown("""
    <style>
        .main-title {
            font-size: 36px;
            color: #4A90E2;
            font-weight: bold;
            text-align: center;
        }
        .section {
            background-color: #f9f9f9;
            padding: 10px;
            border-radius: 10px;
            margin-top: 10px;
        }
        .section p, .section ul {
            color: #666666;
        }
    </style>
""", unsafe_allow_html=True)

@st.cache_resource
def init_spark():
    return sparknlp.start()

@st.cache_resource
def create_pipeline(model):
    imageAssembler = ImageAssembler() \
        .setInputCol("image") \
        .setOutputCol("image_assembler")

    imageCaptioning = VisionEncoderDecoderForImageCaptioning \
        .pretrained("image_captioning_vit_gpt2") \
        .setBeamSize(2) \
        .setDoSample(False) \
        .setInputCols(["image_assembler"]) \
        .setOutputCol("caption")

    pipeline = Pipeline(stages=[imageAssembler, imageCaptioning])
    return pipeline

def fit_data(pipeline, data):
    empty_df = spark.createDataFrame([['']]).toDF('text')
    model = pipeline.fit(empty_df)
    light_pipeline = LightPipeline(model)
    annotations_result = light_pipeline.fullAnnotateImage(data)
    return annotations_result[0]['caption'][0].result

def save_uploadedfile(uploadedfile):
    filepath = os.path.join(IMAGE_FILE_PATH, uploadedfile.name)
    with open(filepath, "wb") as f:
        if hasattr(uploadedfile, 'getbuffer'):
            f.write(uploadedfile.getbuffer())
        else:
            f.write(uploadedfile.read())
        
# Sidebar content
model_list = ['image_captioning_vit_gpt2']
model = st.sidebar.selectbox(
    "Choose the pretrained model",
    model_list,
    help="For more info about the models visit: https://sparknlp.org/models"
)

# Set up the page layout
st.markdown(f'<div class="main-title">VisionEncoderDecoder For Image Captioning</div>', unsafe_allow_html=True)
# st.markdown(f'<div class="section"><p>{sub_title}</p></div>', unsafe_allow_html=True)

# Reference notebook link in sidebar
link = """
<a href="https://colab.research.google.com/github/JohnSnowLabs/spark-nlp/blob/master/examples/python/annotation/image/VisionEncoderDecoderForImageCaptioning.ipynb">
    <img src="https://colab.research.google.com/assets/colab-badge.svg" style="zoom: 1.3" alt="Open In Colab"/>
</a>
"""
st.sidebar.markdown('Reference notebook:')
st.sidebar.markdown(link, unsafe_allow_html=True)

# Load examples
IMAGE_FILE_PATH = f"inputs"
image_files = sorted([file for file in os.listdir(IMAGE_FILE_PATH) if file.split('.')[-1]=='png' or file.split('.')[-1]=='jpg' or file.split('.')[-1]=='JPEG' or file.split('.')[-1]=='jpeg'])

img_options = st.selectbox("Select an image", image_files)
uploadedfile = st.file_uploader("Try it for yourself!")

if uploadedfile:
    file_details = {"FileName":uploadedfile.name,"FileType":uploadedfile.type}
    save_uploadedfile(uploadedfile)
    selected_image = f"{IMAGE_FILE_PATH}/{uploadedfile.name}"
elif img_options:
    selected_image = f"{IMAGE_FILE_PATH}/{img_options}"

st.subheader('Classified Image')

image_size = st.slider('Image Size', 400, 1000, value=400, step = 100)

try:
    st.image(f"{IMAGE_FILE_PATH}/{selected_image}", width=image_size)
except:
    st.image(selected_image, width=image_size)

st.subheader('Classification')

spark = init_spark()
Pipeline = create_pipeline(model)
output = fit_data(Pipeline, selected_image)

st.markdown(f'This document has been classified as  : **{output}**')