|
import numpy as np
|
|
import streamlit as st
|
|
from PIL import Image
|
|
|
|
from models.deep_learning import DeepLearningGradCAM, ImageSimilarity
|
|
from utils import configs
|
|
from utils.functional import (
|
|
generate_empty_space,
|
|
get_default_images,
|
|
get_most_salient_object,
|
|
set_page_config,
|
|
set_seed,
|
|
)
|
|
|
|
|
|
set_seed()
|
|
|
|
|
|
set_page_config("Image Similarity with Deep Learning", "🖼️")
|
|
|
|
|
|
name_model = st.sidebar.selectbox("Select Model", tuple(configs.NAME_MODELS.keys()))
|
|
support_set_method = st.sidebar.selectbox(
|
|
"Select Support Set Method", configs.SUPPORT_SET_METHODS
|
|
)
|
|
freeze_model = st.sidebar.checkbox("Freeze Model", value=True)
|
|
pretrained_model = st.sidebar.checkbox("Pretrained Model", value=True)
|
|
|
|
|
|
|
|
@st.cache_resource
|
|
def load_model(
|
|
name_model: str, support_set_method: str, freeze_model: bool, pretrained_model: bool
|
|
):
|
|
image_similarity = ImageSimilarity(
|
|
name_model, freeze_model, pretrained_model, support_set_method
|
|
)
|
|
custom_grad_cam = DeepLearningGradCAM(
|
|
name_model, freeze_model, pretrained_model, support_set_method
|
|
)
|
|
return image_similarity, custom_grad_cam
|
|
|
|
|
|
image_similarity, custom_grad_cam = load_model(
|
|
name_model, support_set_method, freeze_model, pretrained_model
|
|
)
|
|
|
|
|
|
st.markdown("# ❓ Application Description")
|
|
st.write(
|
|
"""
|
|
Looking for a powerful and efficient way to find similar images? Look no further than our Image Similarity with Deep Learning application! Using the latest in deep learning technology, this app allows you to quickly and easily identify images that are similar in content, style, and more.
|
|
|
|
Whether you're a professional photographer, a graphic designer, or just someone who loves taking pictures, our app can help you find the images you need in a flash. With a user-friendly interface and advanced algorithms, it's never been easier to search through large collections of images and find the ones that best match your needs.
|
|
|
|
So why wait? Give our Image Similarity with Deep Learning app a try today and start exploring the world of images like never before! 🖼️
|
|
"""
|
|
)
|
|
|
|
col1, col2 = st.columns(2)
|
|
uploaded_file1 = col1.file_uploader(
|
|
"Upload image file 1", type=["jpg", "jpeg", "png", "bmp", "tiff"]
|
|
)
|
|
select_default_images1 = col1.selectbox("Select default images 1", get_default_images())
|
|
col1.caption("Default Images 1 will be used if no image is uploaded.")
|
|
select_image_button1 = col1.button("Select Image 1")
|
|
if select_image_button1:
|
|
st.success("Image 1 selected")
|
|
|
|
uploaded_file2 = col2.file_uploader(
|
|
"Upload image file 2", type=["jpg", "jpeg", "png", "bmp", "tiff"]
|
|
)
|
|
select_default_images2 = col2.selectbox("Select default images 2", get_default_images())
|
|
col2.caption("Default Images 2 will be used if no image is uploaded.")
|
|
select_image_button2 = col2.button("Select Image 2")
|
|
if select_image_button2:
|
|
st.success("Image 2 selected")
|
|
|
|
if select_image_button1 and uploaded_file1 is not None:
|
|
image1 = np.array(Image.open(uploaded_file1).convert("RGB"))
|
|
st.session_state["image1"] = image1
|
|
elif select_image_button1 and uploaded_file1 is None:
|
|
image1 = np.array(Image.open(select_default_images1).convert("RGB"))
|
|
st.session_state["image1"] = image1
|
|
|
|
if select_image_button2 and uploaded_file2 is not None:
|
|
image2 = np.array(Image.open(uploaded_file2).convert("RGB"))
|
|
st.session_state["image2"] = image2
|
|
elif select_image_button2 and uploaded_file2 is None:
|
|
image2 = np.array(Image.open(select_default_images2).convert("RGB"))
|
|
st.session_state["image2"] = image2
|
|
|
|
if (
|
|
st.session_state.get("image1") is not None
|
|
and st.session_state.get("image2") is not None
|
|
):
|
|
image1 = st.session_state.get("image1")
|
|
image2 = st.session_state.get("image2")
|
|
col1, col2 = st.columns(2)
|
|
col1.write("## 📸 Preview Image 1")
|
|
col1.image(image1, use_column_width=True)
|
|
col2.write("## 📸 Preview Image 2")
|
|
col2.image(image2, use_column_width=True)
|
|
predict_image_button = st.button("Get Image Similarity")
|
|
generate_empty_space(2)
|
|
if predict_image_button:
|
|
with st.spinner("Getting Image Similarity..."):
|
|
result_similarity = image_similarity.get_similarity(image1, image2)
|
|
result_grad_cam1 = custom_grad_cam.get_grad_cam(image1)
|
|
result_grad_cam2 = custom_grad_cam.get_grad_cam(image2)
|
|
inference_time = result_similarity["inference_time"]
|
|
col1, col2 = st.columns(2)
|
|
col1.write("### 😎 Grad CAM Image 1")
|
|
col1.image(result_grad_cam1, use_column_width=True)
|
|
col2.write("### 😎 Grad CAM Image 2")
|
|
col2.image(result_grad_cam2, use_column_width=True)
|
|
col1, col2 = st.columns(2)
|
|
col1.write("### 🤔 Most Salient Object Image 1")
|
|
col1.image(get_most_salient_object(image1), use_column_width=True)
|
|
col2.write("### 🤔 Most Salient Object Image 2")
|
|
col2.image(get_most_salient_object(image2), use_column_width=True)
|
|
st.write("### 📝 Result")
|
|
st.write(f"Similarity Score: {result_similarity['similarity'] * 100:.2f}%")
|
|
st.write(f"Similarity Label: {result_similarity['result_similarity'].title()}")
|
|
st.write(f"Inference Time: {inference_time:.2f} s")
|
|
st.session_state["image1"] = None
|
|
st.session_state["image2"] = None
|
|
|
