Spaces:
Running
Running
| import streamlit as st | |
| import keras | |
| import numpy as np | |
| from PIL import Image | |
| st.set_page_config(layout="wide") | |
| #title | |
| st.title('Crossing Identifier') | |
| #header | |
| st.header('Choose whether you\'d like to enter a latitude/longitude coordinates, or upload a satellite image.') | |
| state = st.session_state | |
| if "dict_options" not in state: | |
| state.dict_options = {} | |
| if "submitted" not in state: | |
| state.submitted = False | |
| options = ["option1", "option2", "option3", "option4"] | |
| col1, col2, col3 = st.columns(3) | |
| with col1.form("my_form"): | |
| new_country = st.text_input("New country") | |
| submit_button = st.form_submit_button( | |
| label="Add new country", on_click=lambda: state.update(submitted=True) | |
| ) | |
| if state.submitted: | |
| state.dict_options[new_country] = col2.multiselect( | |
| f"Select the options you want for {new_country}", | |
| options, | |
| default=options[:2], | |
| ) | |
| col2.write(state.dict_options) | |
| #divide app into two columns | |
| col1, col2 = st.columns(2) | |
| #load model and initialize image size required by model. uploaded images are resized to indicated size | |
| loaded_model = keras.models.load_model("0.0008-0.92.keras") | |
| img_height = 640 | |
| img_width = 640 | |
| #place to enter | |
| #place to enter coordinates (or upload) and display image | |
| with col1: | |
| enter_coords = st.button("Enter Coordinates") | |
| if enter_coords: | |
| st.write(":smile:") | |
| upload_img = st.button("Upload an Image") | |
| if enter_coords: | |
| st.write("ok") | |
| st.header('Please upload a satellite image, or enter a latitude/longitude pair') | |
| img_buffer = st.file_uploader("Upload a satellite image file (format: .png, .jpeg, or .jpg).",type=['png', 'jpeg', 'jpg']) | |
| if img_buffer is not None: | |
| st.image(img_buffer, use_column_width = True) | |
| #place to display prediction result | |
| with col2: | |
| if img_buffer is not None: | |
| st.header('Result') | |
| img = Image.open(img_buffer).convert("RGB") | |
| img_array = np.array(img) | |
| batch_size = 1 | |
| img_array = np.reshape(img_array,[batch_size,img_height,img_width,3]) | |
| result = loaded_model.predict(img_array) | |
| st.write("Your prediction is:") | |
| st.write(f"{np.round(result[0][0]*100,decimals=2)}% chance of no crossing") | |
| st.write(f"{np.round(result[0][1]*100,decimals=2)}% chance of at least one crossing") | |