streamlit_app.py

import numpy as np
import streamlit as st
import cv2 as cv
from PIL import Image
from keras.models import load_model
from keras.layers import TFSMLayer



# Label traffic signs
labels_dict = {
    0: '274. Speed limit (20km/h)',
    1: '274. Speed limit (30km/h)',
    2: '274. Speed limit (50km/h)',
    3: '274. Speed limit (60km/h)',
    4: '274. Speed limit (70km/h)',
    5: '274. Speed limit (80km/h)',
    6: '278. End of speed limit (80km/h)',
    7: '274. Speed limit (100km/h)',
    8: '274. Speed limit (120km/h)',
    9: '276. No passing',
    10: '277. No passing vehicle over 3.5 tons',
    11: '301. Right-of-way at intersection',
    12: '306. Priority road',
    13: '205. Yield',
    14: '206. Stop',
    15: '250. No vehicles',
    16: '253. Vehicle > 3.5 tons prohibited',
    17: '267. No entry',
    18: '101. General caution',
    19: '104. Dangerous curve left',
    20: '103. Dangerous curve right',
    21: '105-10. Double curve',
    22: '112. Bumpy road',
    23: '114. Slippery road',
    24: '121. Road narrows on the right',
    25: '123. Road work',
    26: '131. Traffic signals',
    27: '133. Pedestrians',
    28: '136. Children crossing',
    29: '138. Bicycles crossing',
    30: '101-51. Beware of ice/snow',
    31: '142. Wild animals crossing',
    32: '282. End speed + passing limits',
    33: '209-20. Turn right ahead',
    34: '209-10. Turn left ahead',
    35: '209-30. Ahead only',
    36: '214-20. Go straight or right',
    37: '214-10. Go straight or left',
    38: '222-20. Keep right',
    39: '222-10. Keep left',
    40: '215. Roundabout mandatory',
    41: '280. End of no passing',
    42: '281. End no passing vehicle > 3.5 tons'
}

# model = tf.keras.models.load_model('keras_model/')
# # model = TFSMLayer('keras_model/', call_endpoint='serving_default')
# model = load_model('keras_model/')''

model = TFSMLayer('keras_model/', call_endpoint='serving_default')

@st.cache(allow_output_mutation=True)
def sign_predict(image):
    image = np.array(image, dtype=np.float32)
    image = image/255
    image = np.reshape(image, (1, 32, 32))
    x = image.astype(np.float32)
    prediction = model.predict(x)
    prediction_max = np.argmax(prediction)
    prediction_label = labels_dict[prediction_max]
    confidence = np.max(prediction)
    return prediction_label, confidence


def main():
    # Set page config and markdowns
    st.set_page_config(page_title='Traffic Signs Classifier', page_icon=':car:')
    st.title('Traffic Signs Classifier')
    st.markdown("""
        This application classifies traffic signs. Upload any photo of a traffic sign 
        and receive its name out of 43 present classes. For getting the correct prediction, 
        try to upload a square picture containing only the sign.
        """)
    with st.expander("See list of classes"):
        st.write(list(labels_dict.values()))
    st.image('./app_images/road_sign.jpg', use_column_width=True)
    image_usr = st.file_uploader('Upload a photo of traffic sign here', type=['jpg', 'jpeg', 'png'])

    if image_usr is not None:
        col1, col2 = st.columns(2)
        col1.markdown('#### Your picture')
        col2.markdown('#### Your picture 32x32 gray')
        image = Image.open(image_usr)
        with col1:
            st.image(image, use_column_width=True)

        image_np = np.array(image.convert('RGB'))
        image_col = cv.cvtColor(image_np, 1)
        image_gray = cv.cvtColor(image_col, cv.COLOR_BGR2GRAY)
        image_32 = cv.resize(image_gray, (32, 32))
        with col2:
            st.image(image_32, use_column_width=True)

        # Make prediction
        prediction_label, confidence = sign_predict(image_32)

        st.write('##### Prediction:', prediction_label)
        st.write('##### Confidence:', str(confidence))
        st.markdown('***')

    # Markdowns
    st.subheader('About this app')
    st.markdown("""
    The app uses an implementation of LeNet-5 Convolutional Neural Network. 
    The model was trained and tested on about 40.000 real photos of 43 types of Indian traffic signs.
    
    Data was taken from The Indian Traffic Sign Recognition :
    https://huggingface.co/datasets/kannanwisen/Indian-Traffic-Sign-Classification
    
    
    Author: Prajwal Thawari
    """)


if __name__ == '__main__':
    main()