Update README.md

examples/imagecaption/README.md

@@ -3,6 +3,7 @@
 This is a partial implementation of "Show and Tell: A Neural Image Caption Generator" (http://arxiv.org/abs/1411.4555), borrowing heavily from Andrej Karpathy's NeuralTalk (https://github.com/karpathy/neuraltalk)
 
 This example consists of three parts:
+
 1. COCO Preprocessing - prepare the dataset by precomputing image representations using GoogLeNet
 2. RNN Training - train a network to predict image captions
-3. Caption Generation - use the trained network to caption new images
+3. Caption Generation - use the trained network to caption new images