keras-flower 0.3

A Simple Flower classification package trained on DenseNet201

Keras Flower

A Simple Flower classification package using DenseNet201.

• Classes : 104 classes of flowers
• Training Data Set : Petals to the Metal - Flower Classification on TPU

Usage

• Use pip install to install this package

pip install keras_flower

• To get all prediction results

import keras_flower as kf
predictions = kf.predict_by_path("file/to/predict.png")
print(predictions)

Sample output:

[1.4414026e-06 1.6031330e-06 1.6295390e-06 1.1156463e-06 2.7592062e-06
...
 1.1587109e-06 4.1556059e-06 1.0784672e-05 6.0254356e-06]

• To get top prediction result with flower labels

import keras_flower as kf
for predicted, score in kf.predict_name_by_path("/path/to/file.png"):
    print(predicted, score)

Sample output:

sunflower 0.99960905

Validation

• Data set : 102 Category Flower Dataset
• No of images : 8189
• Overall accuracy : 0.9715
• confusion matrix : 102flowers_confusion_matrix.csv
• classification report : 102flowers_classification_report.txt
• classification report summary:

                           precision    recall  f1-score   support
                micro avg       0.97      0.97      0.97      8189
                macro avg       0.95      0.94      0.94      8189
             weighted avg       0.97      0.97      0.97      8189

Demo