Skip to content

TF Keras Applications EfficientNetB0

TensorFlow support

The TensorFlow path of the RBLN Compiler is verified with TensorFlow 2.18 and tf_keras 2.18 (the Keras 2 API) on Python 3.10 to 3.12. Keras 3 models are not supported, and later TensorFlow releases are not verified.

Overview

This tutorial shows how to compile and run inference with the TF Keras Applications EfficientNetB0 model for image classification using the RBLN Python API.

Setup & Installation

Before you begin, ensure that your system environment is properly configured and that all required packages are installed. This includes:

  • System Requirements:
  • Package Requirements:
  • Installation Command:
    1
2
3
4
    pip install tensorflow==2.18.0 tf_keras==2.18.0 numpy pillow
pip install \
  --extra-index-url https://pypi.rbln.ai/simple \
  rebel-compiler==0.12.0

Info

  • rebel-compiler requires an RBLN Portal account.
  • The commands above are intended for a default pip install on Debian-based Linux such as Ubuntu. For all other configurations, refer to the Installation Guide for the supported install matrix and the applicable commands.

Using RBLN Python API

Model Compilation

Import EfficientNetB0 from the TF Keras Applications module, instantiate the model with pre-trained weights, and convert it into a tf.function.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
import tensorflow as tf  
from tf_keras.applications.efficientnet import EfficientNetB0  
import rebel  # RBLN Compiler  

# Instantiate TF Keras Applications EfficientNetB0 model  
model = EfficientNetB0(weights='imagenet')  
func = tf.function(lambda input_img: model(input_img))  

# Compile the model  
input_info = [('input_img', [1, 224, 224, 3], tf.float32)]  
compiled_model = rebel.compile_from_tf_function(  
    func,  
    input_info,  
)  

# Save the compiled model to disk  
compiled_model.save('efficientnetb0.rbln')

Model Inference

Download and preprocess the input image required for EfficientNetB0. Use preprocess_input() for proper preprocessing. Load the compiled model using RBLN Runtime and run inference on the preprocessed input. Decode the output predictions using decode_predictions() and display the top prediction.

 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
import urllib.request  
from tf_keras.preprocessing import image  
from tf_keras.applications.efficientnet import preprocess_input, decode_predictions  
import numpy as np  

import rebel  # RBLN Runtime

# Prepare the input  
img_url = 'https://rbln-public.s3.ap-northeast-2.amazonaws.com/images/tabby.jpg'  
img_path = './tabby.jpg'  
with urllib.request.urlopen(img_url) as response, open(img_path, 'wb') as f:  
    f.write(response.read())  
img = image.load_img(img_path, target_size=(224, 224))  
x = image.img_to_array(img)  
x = np.expand_dims(x, axis=0)  
x = preprocess_input(x)  
x = np.ascontiguousarray(x)  

# Load the compiled model  
module = rebel.Runtime('efficientnetb0.rbln')  

# Run inference  
rebel_result = module.run(x)  

# Check results  
print('Top1 category:', decode_predictions(rebel_result, top=1)[0])

Example Output: 

1
Top1 category: [('n02123045', 'tabby', 0.3798828)]

References