Destroy Image Classification

Ensemble of Pre-trained models for Image Classification

Prerequisite

Python 3.X
Tensorflow 2.X
Numpy
Pandas

Pre_trained Models

• Xception : Documentation
• MobilnetV2 : Documentation
• InceptionV3 : Documentation

Steps

1. Load data and Fine tune the pre-trained models
2. Create your customized data genertor
you need to build your own data generator to feed the pre-trained models simultaniously in this exemple i used three pre-trained models so i create a data generator that generate 3 copies of data for training and one label. So it's depend on the number of pre-trained models to use
3. Build the combinaison
First of all we need to define our input and output of our model then we will use the functional API Concatenate the models and the dense layers

GeM Pooling + ArcFace

GeM
Arcface