Read-me.txt

Since this course was originally release, the transfer learning code was removed from the TensorFlow source examples
referenced in the course.  In addition, some users had issues working with zipped files containing the 
flower images used to retrain the model.  To rectify both of these issues the following are contained in 
this folder.

retrain.py -  The program used to retrain the inception model to classify flowers.  This is example shown 
in the course.

flower_photos - a folder containing the uncompressed images used to retrain the model.  There are 5 folders 
containing images, labelled by the flower class (daisy, dandelion, roses, sunflowers, tulips).

To retrain the model using theimages in the flower_photos folder:
- Ensure you are in the folder containing retrain.py
- Enter the command: 

      python retrain.py --image_dir ./flower_photos

The retraining may take a while, so be patient.  

By default, the retrained model is the file /tmp/output_graph.pb.  And the labels, flower types, found in 
in the file /tmp/output_labels.txt.  

You can test the retrained model on flower imaged by selecting an image from the test-images folder and running
program try-retrain.py to see how well the model predicts the class of the image. For example, if use the command:

     python try-retrain.py test-images\Daisy1.jpg

You should see an output like the one shown below.

Image Classification Probabilities
daisy (score = 0.99684)
dandelion (score = 0.00137)
sunflowers (score = 0.00127)
tulips (score = 0.00031)
roses (score = 0.00022)

The output above shows that the retrained model produces a 99.684% probability that the image in the file 
test-images\Daisy1.jpg is in the class "daisy".  

Have fun and try some of the other files in the test-images folder.  Some of them make some interesting 
class predictions!