Image Classification

Flower Recognition - Performance Evaluation on CNN

Trained the system for Flower Recognition using Convolutional Neural Network. A performance evaluation was done on the architecture design. First trained the system with only one hidden layer. Then trained on two hidden layers. The accuracy graph is shown below.

Database

The dataset for Flower Recognition was downloaded from Kaggle.

Image Size = 320X240
Total Images: 4242
Number of Classes: 5

Class Labels:

* chamomile 
* tulip 
* rose
* sunflower 
* dandelion

Implementation

• Python 3.6.2
• Numpy
• TFLearn

Screenshots of Performance

Accuracy: Epoch_20 - ConvLayer_1

Accuracy: Epoch_40 - ConvLayer_2