Add chapter 4 figures

Author: meherkasam

README.md

@@ -44,7 +44,7 @@ We delve into the world of image classification in a mere five lines of Keras co
 
 We use transfer learning to reuse a previously trained network on a new custom classification task to get near state-of-the-art accuracy in a matter of minutes. We then slice and dice the results to understand how well is it classifying. Along the way, we build a common machine learning pipeline, which is repurposed throughout the book. Bonus: we hear from **Jeremy Howard**, co-founder of fast.ai, on how hundreds of thousands of students use transfer learning to jumpstart their AI journey.
 
-[**Chapter 4 - Building a Reverse Image Search Engine: Understanding Embeddings**](https://github.com/practicaldl/Practical-Deep-Learning-Book/tree/master/code/chapter-4) | [Read online](https://learning.oreilly.com/library/view/practical-deep-learning/9781492034858/ch04.html)
+[**Chapter 4 - Building a Reverse Image Search Engine: Understanding Embeddings**](https://github.com/practicaldl/Practical-Deep-Learning-Book/tree/master/code/chapter-4) | [Read online](https://learning.oreilly.com/library/view/practical-deep-learning/9781492034858/ch04.html) | [Figures](figures/chapter-4)
 
 Like Google Reverse Image Search, we explore how one can use embeddings—a contextual representation of an image to find similar images in under ten lines. And then the fun starts when we explore different strategies and algorithms to speed this up at scale, from thousands to several million images, and making them searchable in microseconds.
 

figures/README.md

@@ -11,7 +11,7 @@ This directory contains references to the figures used within the book. All imag
 | [Chapter 1 - Exploring the Landscape of Artificial Intelligence](chapter-1/) |
 | [Chapter 2 - What’s in the Picture: Image Classification with Keras](chapter-2/) |
 | [Chapter 3 - Cats versus Dogs: Transfer Learning in 30 Lines with Keras](chapter-3/) |
-| Chapter 4 - Building a Reverse Image Search Engine: Understanding Embeddings |
+| [Chapter 4 - Building a Reverse Image Search Engine: Understanding Embeddings](chapter-4/) |
 | Chapter 5 - From Novice to Master Predictor: Maximizing Convolutional Neural Network Accuracy |
 | Chapter 6 - Maximizing Speed and Performance of TensorFlow: A Handy Checklist |
 | Chapter 7 - Practical Tools, Tips, and Tricks |

figures/chapter-1/README.md

@@ -1,4 +1,4 @@
-# Chapter 1: Exploring the Landscape of Artificial Intelligence
+# Chapter 1 - Exploring the Landscape of Artificial Intelligence
 
 Note: All images in this directory, unless specified otherwise, are licensed under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/legalcode).
 

figures/chapter-2/README.md

@@ -1,4 +1,4 @@
-# Chapter 2: What's in the Picture: Image Classification with Keras
+# Chapter 2 - What's in the Picture: Image Classification with Keras
 
 Note: All images in this directory, unless specified otherwise, are licensed under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/legalcode).
 

figures/chapter-3/README.md

@@ -1,4 +1,4 @@
-# Chapter 3: Cats Versus Dogs: Transfer Learning in 30 Lines with Keras
+# Chapter 3 - Cats Versus Dogs: Transfer Learning in 30 Lines with Keras
 
 Note: All images in this directory, unless specified otherwise, are licensed under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/legalcode).
 
@@ -13,7 +13,7 @@ Note: All images in this directory, unless specified otherwise, are licensed und
 | [3-5](5-finetuning.png?raw=true) | Fine tuning a convolutional neural network | |
 | [3-6](6-folder-tree-structure.png?raw=true) | Example directory structure of the training and validation data for different classes | |
 | [3-7](7-overfitting-underfitting.png?raw=true) | Underfitting, overfitting, and ideal fitting for points close to a sine curve | |
-| [3-8](8-rabbit-augmentations.png?raw=true) | Possible image augmentations generated from a single image | |
+| [3-8](8-rabbit-augmentations.jpg?raw=true) | Possible image augmentations generated from a single image | |
 | [3-9](9-highest-probability-dogs.png?raw=true) | Images with the highest probability of containing dogs | |
 | [3-10](10-lowest-probability-dogs.png?raw=true) | Images with the lowest probability of containing dogs | |
 | [3-11](11-cats-highest-probability-containing-dogs.png?raw=true) | Images of cats with the highest probability of containing dogs | |

figures/chapter-4/1-rgb-histogram.png

@@ -0,0 +1,32 @@
+# Chapter 4 - Building a Reverse Image Search Engine: Understanding Embeddings
+
+Note: All images in this directory, unless specified otherwise, are licensed under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/legalcode).
+
+## Figure List
+
+| Figure number | Description | Notes |
+|:---|:---|:---|
+| [4-1](1-rgb-histogram?raw=true) | RGB histogram-based "Similar Image Detector" program | |
+| [4-2](2-amazon-scan.png?raw=true) | Product scanner in Amazon app with visual features highlighted | |
+| [4-3](3-tqdm.png) | Progress bar shown with tqdm_notebook | |
+| [4-4](4-original-image.png?raw=true) |The query image from the Caltech-101 dataset | |
+| [4-5](5-first-result.png?raw=true) | The nearest neighbor to our query image | |
+| [4-6](6-second-closest.png?raw=true) | The second nearest neighbour of the queried image | |
+| [4-7](7-nearest-neighbors.png?raw=true) | Nearest neighbor for different images returns similar-looking images | |
+| [4-8](8-tsne.png?raw=true) | t-SNE visualizing clusters of image features, where each cluster represents one object class in the same color | |
+| [4-9](9-tsne-clusters.png?raw=true) | t-SNE visualization showing image clusters; similar images in the same cluster | |
+| [4-10](10-tsne-grid.png?raw=true) | t-SNE visualization with tiled images; similar images are close together | |
+| [4-11](11-tensorflow-embedding-projector.png?raw=true) | TensorFlow Embedding projector showing a 3D representation of 10,000 common English words and highlighting words related to "Beatles" | |
+| [4-12](12-variance-vs-num-pca-dimensions.png?raw=true) | Variance for each PCA dimension |
+| [4-13](13-cumulative-variance-vs-num-pca-dimensions.png?raw=true) | Cumulative variance with each PCA dimension | |
+| [4-14](14-test-time-vs-accuracy.png?raw=true) | Test time versus accuracy for each PCA dimension | |
+| [4-15](15-recall-vs-qps.png?raw=true) | Comparison of ANN libraries ([data source](http://ann-benchmarks.com/)) | |
+| [4-16](16-before-finetune.png?raw=true) | t-SNE visualization of feature vectors of least-accurate classes before fine tuning | |
+| [4-17](17-after-finetune.png?raw=true) | t-SNE visualization of feature vectors of least-accurate classes after fine tuning | |
+| [4-18](18-siamese-network-flowchart.png?raw=true) | A Siamese network for signature verification; note that the same CNN was used for both input images | |
+| [4-19](https://code.flickr.net/2017/03/07/introducing-similarity-search-at-flickr/) | Similar patterns of a desert photo | |
+| [4-20](https://labs.pinterest.com/user/themes/pin_labs/assets/paper/visual_search_at_pinterest.pdf) | The Similar Looks feature of the Pinterest application | |
+| [4-21](21-celebs-like-me.png?raw=true) | Testing our friend Pete Warden’s photo on the [celebslike.me](celebslike.me) website | |
+| [4-22](https://papers.nips.cc/paper/5004-deep-content-based-music-recommendation.pdf) | t-SNE visualization of the distribution of predicted usage patterns, using latent factors predicted from audio | Page 7 |
+| [4-23a](23a-image-captioning.png?raw=true), [4-23b](23b-image-captioning.png?raw=true), [4-23c](23c-image-captioning.png?raw=true) | Image captioning feature in Seeing AI: the Talking Camera App for the blind community | |
+| [4-24](https://arxiv.org/pdf/1608.08716.pdf) | Defining a CNN and visualizing the output of each layer during training in ConvNetJS | Page 2 |