app.py

import torch
import streamlit as st
import numpy as np
from PIL import Image, ImageDraw
from transformers import pipeline
from tempfile import NamedTemporaryFile

imagepipe = pipeline("image-classification", model="flatmoon102/fruits_and_vegetables_image_classification")
detector = pipeline(model="google/owlvit-base-patch32", task="zero-shot-object-detection")

uploaded_image_file = st.file_uploader("Choose an image file")
if uploaded_image_file is not None:
    with NamedTemporaryFile() as temp:
        temp.write(uploaded_image_file.getvalue())
        temp.seek(0)
        result = imagepipe(temp.name)
        st.write(result)

st.title('Upload an image file to detection')

uploaded_image_zero_file = st.file_uploader("Choose an image file (zero)")
texts = st.text_input('tags')

if uploaded_image_zero_file is not None:
    image = Image.open(uploaded_image_zero_file)
    #temp.write(uploaded_image_file.getvalue())
    #temp.seek(0)
    outputImage = np.array(image)
    predictions = detector(
        image,
        candidate_labels=['eggs', 'apple', 'pear']
    )
    st.image(outputImage)

if st.button('apply tag'):
    tags = [['eggs', 'apple', 'pear']]
    #inputs = processor(text=tags, images=image, return_tensors="pt")
    #outputs = model(**inputs)
    #target_sizes = torch.Tensor([image.size[::-1]])
    #results = processor.post_process_object_detection(outputs=outputs, threshold=0.1, target_sizes=target_sizes)

    draw = ImageDraw.Draw(image)

    for prediction in predictions:
        box = prediction["box"]
        label = prediction["label"]
        score = prediction["score"]

        xmin, ymin, xmax, ymax = box.values()
        draw.rectangle((xmin, ymin, xmax, ymax), outline="red", width=3)
        draw.text((xmin, ymin), f"{label}: {round(score,5)}", fill="white")

    st.image(image)
    st.write(predictions)