Update README.md

Author: Robert Coleman

README.md

@@ -5,6 +5,12 @@ Classifies an image as containing either a dog or a cat (using Kaggle's <a href=
 To run these scripts/notebooks, you must have keras, numpy, scipy, and h5py installed, and enabling GPU acceleration is highly recommended if that's an option.
 
 ## img_clf.py
+After playing around with hyperparameters a bit, this reaches around 96-98% accuracy on the validation data, and when tested on Kaggle's hidden test data achieved a log loss score around 0.18.
+
+Most of the code / strategy here was based on <a href="https://blog.keras.io/building-powerful-image-classification-models-using-very-little-data.html">this</a> Keras tutorial.
+
+Pre-trained VGG16 model weights can be downloaded <a href="https://gist.github.com/baraldilorenzo/07d7802847aaad0a35d3">here</a>.
+
 The data directory structure I used was:
 
 * project
@@ -17,12 +23,6 @@ The data directory structure I used was:
       * cats
     * test
       * test
-    
-After playing around with hyperparameters a bit, this reaches around 96-98% accuracy on the validation data, and when tested in Kaggle's competition maxed out with a log loss score around 0.18.
-
-Most of the code / strategy here was based on <a href="https://blog.keras.io/building-powerful-image-classification-models-using-very-little-data.html">this</a> Keras tutorial.
-
-Pre-trained VGG16 model weights can be downloaded <a href="https://gist.github.com/baraldilorenzo/07d7802847aaad0a35d3">here</a>.
 
 ## cats_n_dogs.ipynb:
 This produced a slightly better score (.161 log loss on kaggle test set). The better score most likely comes from having larger images and ensembling a few models, despite the fact there's no image augmentation in the notebook. 
@@ -37,4 +37,20 @@ Might run into memory errors because of the large image dimensions -- if so redu
     * test
       * test
 
+## cats_n_dogs_BN.ipynb:
+This produced the best score (0.069 loss without any ensembling). The notebook incorporates some of the techniques from Jeremy Howard's <a href="http://course.fast.ai/">deep learning class</a> , with the inclusion of batch normalization being the biggest factor. I also added extra layers of augmentation to the prediction script, which greatly improved performance.
+
+Pre-trained model weights for VGG16 w/ batch normalization can be downloaded <a href="http://www.platform.ai/models/">here</a>.
+
+The VGG16BN class is defined in <em>vgg_bn.py</em>, and the data directory structure used was: 
 
+* project
+  * data
+    * train
+      * dogs
+      * cats
+    * validation
+      * dogs
+      * cats
+    * test
+      * test