inference.py

# !wget https://github.com/EliSchwartz/imagenet-sample-images/raw/master/n02085936_Maltese_dog.JPEG

large_img = np.array(Image.open("n02085936_Maltese_dog.JPEG"))
print(f"Image shape: {large_img.shape}")
plt.imshow(large_img)

# Output:
# Image shape: (500, 375, 3)

sess = rt.InferenceSession("combined_dynamic_input.onnx") # Start the inference session and open the model
input_image = np.expand_dims(large_img.astype(np.float32), 0) # Use the input_example from block 0 as input
output = sess.run(["softmaxout_1"], {"X": input_image}) # Compute the standardized output
print("Check:")
print(np.shape(output))

print("Index ", np.argmax(output))
print("Confidence ", np.max(output))

# Output:
# Check:
# (1, 1, 1000, 1, 1)
# Index  153
# Confidence  0.9580029

# According to imagenet index to text list https://gist.github.com/yrevar/942d3a0ac09ec9e5eb3a we can see that 153 stands for 'Maltese dog, Maltese terrier, Maltese'