Docs: Mention batch-represent with the classification demo.

docs/demo-3-classifier.md

@@ -43,8 +43,12 @@ Change `8` to however many
 separate processes you want to run:
 `for N in {1..8}; do ./util/align-dlib.py <path-to-raw-data> align affine <path-to-aligned-data> --size 96 &; done`.
 
-## 3. Create the Classification Model
-Use `./demos/classifier.py train <path-to-aligned-data>` to produce
+## 3. Generate Representations
+`./batch-represent/main.lua -outDir <feature-directory> -data <path-to-aligned-data>`
+creates `reps.csv` and `labels.csv` in `<feature-directory>`.
+
+## 4. Create the Classification Model
+Use `./demos/classifier.py train <feature-directory>` to produce
 the classification model which is an SVM saved to disk as
 a Python pickle.
 

docs/training-new-models.md

@@ -8,12 +8,11 @@ There is a distinction between training the DNN model for feature representation
 and training a model for classifying people with the DNN model.
 If you're interested in creating a new classifier,
 see [Demo 3](http://cmusatyalab.github.io/openface/demo-3-classifier/).
+This page is for advanced users interested in training a new DNN model
+and should be done with large datasets (>500k images) to improve the
+feature representation.
 
-Training a new DNN model is for advanced users and should be done
-with large datasets (>500k images) to improve the feature representation,
-not for classification.
-
-Warning: Training is computationally and memory expensive and takes a
+*Warning:* Training is computationally and memory expensive and takes a
 few weeks on our Tesla K40 GPU.
 Because of this, the training code assumes CUDA is installed.