Comparison of MobileNet, ResNet50, and InceptionV3 in Keras

This is a simple example of using ResNet, MobileNet and InceptionV3 from Keras to do object detection and classification tasks.

Demo

In this case, both MobileNet and InceptionV3 successfully recognize the traffic light, but the ResNet50 seems to be confused by it. Also, as for Model size, we can see from the Model Comparision section, and get to know that MobileNet has really small size. MobileNet can be a great choice for real-time applications.

Papers

• MobileNet: Efficient Convolutional Neural Networks for Mobile Vision Applications

• ResNet50: Deep Residual Learning for Image Recognition

• InceptionV3:Rethinking the Inception Architecture for Computer Vision

1. Model Comparison

2. How to run the code:

1. Select the network by changing the choice value in main.py:

# 1: MobileNet, 2: ResNet50, 3: InceptionV3choice = 1

2. Run the main.py:

python  main.py

3. Requirements:

• keras>=2.0.8

• tensorflow>=1.3.0 or tensorflow-gpu>=1.3.0

• numpy>=1.13.1

• python >= 3.5