Training The Image classifer cats dogs with one more class

raviteja84 · April 22, 2019, 5:59am

I am using the following code:

llSourcell/how_to_make_an_image_classifier/blob/master/demo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Building an image classification model using very little data  \n",
    "\n",
    "Based on the tutorial by Francois Chollet @fchollet https://blog.keras.io/building-powerful-image-classification-models-using-very-little-data.html and the workbook by Guillaume Dominici https://github.com/gggdominici/keras-workshop\n",
    "\n",
    "This tutorial presents several ways to build an image classifier using keras from just a few hundred or thousand pictures from each class you want to be able to recognize.\n",
    "\n",
    "We will go over the following options:  \n",
    "\n",
    "- training a small network from scratch (as a baseline)  \n",
    "- using the bottleneck features of a pre-trained network  \n",
    "- fine-tuning the top layers of a pre-trained network  \n",
    "  \n",
    "This will lead us to cover the following Keras features:   \n",
    "  \n",

This file has been truncated. show original

Where he does a simple implementation of cats vs dogs. I have used my own imageset which consists of just 240 training images and 50 test samples and I have tried this same model. The only thing I have changed in the model is this

img_width, img_height = 432, 288

nb_train_samples = 480
nb_validation_samples = 100

For two classes good and bad in 5 epochs its reaching 100% validation accuracy. When i add a third class into this for Model/Train/Mixed and Model/Validation/Mixed. Then its giving me a a steady accuracy of 33% only and no improvement from there. Where am i going wrong?

For the code:
https://colab.research.google.com/drive/1iMJCLsQDC73obejWsdAxygpvGAWMb2Gw