File size: 961 Bytes
feeed5b 9478d60 feeed5b 9478d60 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 |
import numpy as np
import gradio as gr
import tensorflow as tf
from io import StringIO
from PIL import Image
labels = []
model = tf.keras.models.load_model('./models.h5')
with open("labels.txt") as f:
for line in f:
labels.append(line.replace('\n', ''))
def classify_image(inp):
# Create a copy of the input array to avoid reference issues
inp_copy = np.copy(inp)
# Resize the input image to the expected shape (224, 224)
inp_copy = Image.fromarray(inp_copy)
inp_copy = inp_copy.resize((224, 224))
inp_copy = np.array(inp_copy)
inp_copy = inp_copy.reshape((-1, 224, 224, 3))
inp_copy = tf.keras.applications.efficientnet.preprocess_input(inp_copy)
prediction = model.predict(inp_copy).flatten()
confidences = {labels[i]: float(prediction[i]) for i in range(90)}
return confidences
demo = gr.Interface(classify_image, gr.Image(), gr.Label(num_top_classes=3))
if __name__ == "__main__":
demo.launch() |