Image Recognition using MobileNet

Introduction

The process of identifying an object or feature with an image is known as Image Recognition. Image recognition finds its place in diverse domains be it Medical imaging, automobiles, security, or detecting defects.

What is MobileNet and Why is it so Popular?

MobileNet is deep learning CNN model developed using depth−wise separable convolutions. This model highly decreases the number of parameters when compared to other models of the same depth. This model is lightweight and is optimized to run on mobile and edge devices. There are three versions of Mobilenet released so far.ie MobileNet v1, v2 and v3. Mobilenet is developed by Google.

Let's talk a bit about MobileNet V1 and V2 which have been in the ML space for quite some time now.

MobileNetV2 provides two extensive features over MobileNet V1. They are

• MobileNetV2 has a linear bottleneck between the layers. It preserves information by not allowing non−linearities to destroy too much information

• Short connections between bottlenecks

Architecture of Mobilenetv2

A comparison between MobileNet v1 and Mobilenet V2

Code Implementation of Image Recognition

Example

## MOBILENET

import numpy as np
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras.layers import Dense, Activation
from tensorflow.keras.metrics import categorical_crossentropy
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.preprocessing import image
from tensorflow.keras.models import Model
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.applications import imagenet_utils
import matplotlib.pyplot as plt
from IPython.display import Image,display
%matplotlib inline

mobile = tf.keras.applications.mobilenet.MobileNet()

def format_image(file):
  image_path = '/content/images/'
  img = image.load_img(image_path + file, target_size=(224, 224))
  img_array = image.img_to_array(img)
  img_array_exp_dims = np.expand_dims(img_array, axis=0)
  return tf.keras.applications.mobilenet.preprocess_input(img_array_exp_dims)


display(Image(filename='/content/images/image.jpg', width=300,height=200))

preprocessed_img = format_image('image.jpg')
prediction_results = mobile.predict(preprocessed_img)

results = imagenet_utils.decode_predictions(prediction_results)
print(results)

Output

[[('n02279972', 'monarch', 0.58884907), ('n02281406', 'sulphur_butterfly', 
0.18508224), ('n02277742', 'ringlet', 0.15471826), ('n02281787', 'lycaenid', 0.04744451), 
('n02276258', 'admiral', 0.01013135)]]

Advantages of MobileNet over other networks

• MobileNets have higher classification accuracy and fewer parameters.

• MobileNets are small and have low latency with optimized power consumption most suitable for mobile and embedded devices.

• They are highly effective feature extractors for segmentation and object detection

Benefits of Image Recognition

• Used by autonomous vehicles and robots for detecting obstacles

• Highly used in OCR techniques for retrieving information from images

• Traffic lane detection

• Face Detection and attendance systems

• Captioning and tagging images, useful in social media sites.

Conclusion

Image Recognition has become the preliminary task of every object detection and video−related task. With the plethora of pre−trained models and architectures already available, it has become very relevant in the current AI space related to Vision.

Mithilesh Pradhan

Updated on: 30-Dec-2022

276 Views

Kickstart Your Career

Get certified by completing the course

Get Started