Image Classification

Pre-trained VGG-Net Model for image classification using tensorflow

DataSets :

we used each of this DataSets for Image Classification training

• UC Merced Land Dataset
• SIRI-WHU
• RSSCN7

After Training :

Resultat of UC Merced Land DataSet After Image Classification Training

Testing the classification of one batch of Pictures from UC Merced Land Use Dataset

Cost and Accuracy :

graph represent the values of both of cost and accuracy each epoch

How To use :

you can use this model to classify any DataSet just follow the 4 next instruction

Instalation :

• install tensorflow 1.6 matplotlib opencv imutils

 pip install tensorflow matplotlib opencv-python imutils

• to install tensorflow gpu matplotlib opencv

Train the Model:

To Train Model for different DataSets or Different Classification follow the steps :

Exploit your DataSet

python dataSetGenerator.py  [-h] --path path [--SaveTo SaveTo] [--resize resize]
                      [--resize_to resize_to] [--percentage percentage]
                      [--dataAug dataAugmentation]

Example

python dataSetGenerator.py  --path Desktop/SIRI-WHU --resize --resize_to 200

Help

image dataSet as numpy file.

       picture dataSets
          |
          |----------class-1
          |        .   |-------image-1
          |        .   |         .
          |        .   |         .
          |        .   |         .
          |        .   |-------image-n
          |        .
          |-------class-n

optional arguments:
  -h, --help            show this help message and exit
  --path path           the path for picture dataSets folder (/)
  --SaveTo SaveTo       the path when we save dataSet (/)
  --resize resize       choose resize the pictures or not
  --resize_to resize_to
                        the new size of pictures
  --percentage percentage
                        how many pictures you want to use for training
  --dataAug dataAugmentation
                        apply data Augmentation Strategy

Train your DataSet

python   train_vgg19.py [-h] --dataset dataset [--batch batch] [--epochs epochs]

Example

python train_vgg19.py  --dataset SIRI-WHU

Help

Train vgg19 [-h] --dataset dataset [--batch batch] [--epochs epochs]

Simple tester for the vgg19_trainable

optional arguments:
  -h, --help         show this help message and exit
  --dataset dataset  DataSet Name
  --batch batch      batch size
  --epochs epochs    number of epoch to train the network

Test your Model :

to test your model

python test_vgg19.py [-h] --dataset dataset [--batch batch]

Example

python test_vgg19.py  --dataset SIRI-WHU

Help

tester for the vgg19_trainable

optional arguments:
  -h, --help         show this help message and exit
  --dataset dataset  DataSet Name
  --batch batch      batch size

Confusion Matrix :

to Draw Confusion matrix (the output in images)

python confusion_matrix.py -h [-h] --dataset dataset [--batch batch] [--showPic showPic]

Help

Draw Confusion Matrix for the vgg19

optional arguments:
  -h, --help         show this help message and exit
  --dataset dataset  DataSet Name
  --batch batch      batch size
  --showPic showPic  Show patch of picture each epoch

For Distributed Tensorflow [optional] :

1. Download and install nmap

2. install nmap python module

 pip install python-nmap

3. Set Workers and pss (parameter servers) devices name in train_vgg19_distibuted

workers = ['PC1','PC2']
pss = ['PC3']