Index
A
- Accelerate / A brief introduction to Core ML
- action shots / Data to drive the desired effect – action shots
- anchor boxes / Object localization and object detection
- Artificial Intelligence (AI) / What is machine learning?
- assisted drawing
- about / Assisted drawing
- data input / Input data and preprocessing
- preprocessing functionality / Input data and preprocessing
- Augmented Intelligence (AI) / Shadow draw – real-time user guidance for freehand drawing
- augmented reality (AR) / Inference on the edge
- AVFoundation framework / Capturing data
B
- backpropagation / Transferring style from one image to another
- batches
- optimizing with / Optimizing with batches
C
- Cambridge-driving Labeled Video Database (CamVid)
- reference / Data to drive the desired effect – action shots
- class activation maps (CAMs) / Classifying pixels
- classifier / What is machine learning?, Shadow draw – real-time user guidance for freehand drawing
- clustering / Netflix – making recommendations
- Collaborative Filtering (CF) algorithm / Netflix – making recommendations
- Common Objects in Context (COCO) dataset
- conditional random fields (CRF) / Post-processing and ensemble techniques
- confidence value / Object localization and object detection
- convolutional neural network (CNN) / Shutterstock – image search based on composition
- Core ML
- about / A brief introduction to Core ML
- workflow / Workflow
- Keras Tiny YOLO, converting to / Converting Keras Tiny YOLO to Core ML
- Keras model, converting to / Converting a Keras model to Core ML
- Core ML Tools / Workflow
- CoreVideo / Capturing data
- custom layers
- building, in Swift / Building custom layers in Swift
D
- darknet
- reference / Converting Keras Tiny YOLO to Core ML
- data
- inputting / Input data and preprocessing
- preparing / Preparing the data
- data augmentation / Improving the model
- data preprocessing / Shadow draw – real-time user guidance for freehand drawing
- dice coefficient / Data to drive the desired effect – action shots
- DragonBot / Bringing it all together
- drawing functionality
- implementing / Drawing
- dynamic target resizing / iOS keyboard prediction – next letter prediction
E
- edge computing / Inference on the edge
- edge detection / Understanding images
- edge filters / Understanding images
- embeddings / iOS keyboard prediction – next letter prediction
- Euclidean distance / Netflix – making recommendations
- expected output / Transferring style from one image to another
F
- FacialEmotionDetection / Bringing it all together
- facial expressions / Facial expressions
- feature engineering / Shadow draw – real-time user guidance for freehand drawing, Understanding images
- frames per second (fps) / Capturing data
G
- GPU
- using / Taking advantage of the GPU
- gram matrix / Transferring style from one image to another
H
- Hey Siri / Inference on the edge
- histogram of oriented gradients (HOG) / Shadow draw – real-time user guidance for freehand drawing
- human-computer interaction (HCI) / Shadow draw – real-time user guidance for freehand drawing
I
- IBM Watson Services for Core ML / Closing thoughts
- ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) / Shutterstock – image search based on composition
- images
- about / Understanding images
- style, transferring / Transferring style from one image to another
- inference
- versus training / Difference between training and inference
- performing / Performing inference
- International Conference on Machine Learning (ICML) / Facial expressions
- Internet of Things (IoT) / Inference on the edge
- intersection over union (IoU) / Object localization and object detection
- Intersection over Union (IoU) coefficient / Data to drive the desired effect – action shots
- iOS keyboard prediction / iOS keyboard prediction – next letter prediction
K
- K-means / Shadow draw – real-time user guidance for freehand drawing
- Keras model
- converting, to Core ML / Converting a Keras model to Core ML
- Keras Tiny YOLO
- converting, to Core ML / Converting Keras Tiny YOLO to Core ML
- kernels
- stride value / Understanding images
- padding / Understanding images
L
- Labeled Faces in the Wild (LFW)
- reference / Data to drive the desired effect – action shots
- layers
- accelerating / Accelerating our layers
- learning algorithms
- about / Learning algorithms
- auto insurance in Sweden / Auto insurance in Sweden
- supported learning algorithms / Supported learning algorithms
- Long Short-Term Memory (LSTM) / Recurrent Neural Networks for drawing classification
- loss function / Difference between training and inference
M
- machine learning (ML)
- about / What is machine learning?
- use cases / Inference on the edge
- metal performance shaders (MPSes) / A brief introduction to Core ML
- ML algorithms
- about / A brief tour of ML algorithms
- Collaborative Filtering (CF) / Netflix – making recommendations
- shadow draw / Shadow draw – real-time user guidance for freehand drawing
- Shutterstock / Shutterstock – image search based on composition
- iOS keyboard prediction / iOS keyboard prediction – next letter prediction
- ML Kit / Closing thoughts
- ML workflow / A typical ML workflow
- model
- weight, reducing / Reducing your model's weight
- creating / Creating and training a model
- training / Creating and training a model
- model metadata / Model metadata
- model parameters / Model parameters
- multi-class classification / Preparing the data
N
- named entity recognition (NER) / A brief introduction to Core ML
- natural language processing (NLP) / A brief introduction to Core ML
- NDJSON
- Netflix / Netflix – making recommendations
- non-max suppression / Object localization and object detection
O
- object detection / Object localization and object detection
- object localization / Object localization and object detection
- object recognition
- about / Recognizing objects in the world, Object localization and object detection
- data, capturing / Capturing data
- data, preprocessing / Preprocessing the data
- object segmentation
- one-hot encoding / iOS keyboard prediction – next letter prediction
P
- PASCAL VOC
- reference / Data to drive the desired effect – action shots
- photo effects application
- building / Building the photo effects application
- photos
- finding / Making it easier to find photos
- pixels
- classifying / Classifying pixels
- predicted output / Transferring style from one image to another
- preprocessing
- implementing / Input data and preprocessing
- probabilistic results
- working with / Working with probabilistic results
- model, improving / Improving the model
- designing, in constraints / Designing in constraints
- heuristics / Embedding heuristics
- ensemble techniques / Post-processing and ensemble techniques
- post-processing technique / Post-processing and ensemble techniques
- human assistance / Human assistance
Q
- Quick, Draw! / Bringing it all together
R
- Ramer-Douglas-Peucker algorithm
- reference / Input data and preprocessing
- recommendations / Netflix – making recommendations
- Recurrent Neural Networks
- for drawing classification / Recurrent Neural Networks for drawing classification
S
- Sequence to Sequence (Seq2Seq) model / Recurrent Neural Networks for drawing classification
- shadow draw / Shadow draw – real-time user guidance for freehand drawing
- Shutterstock / Shutterstock – image search based on composition
- single instruction, multiple data (SIMD) / Sorting by visual similarity
- Single Instruction, Multiple Data (SIMD) / Accelerating our layers
- sketches
- classifying / Classifying sketches
- sketching functionality / Drawing
- sliding window detection algorithm / Object localization and object detection
- sorting by visual similarity / Sorting by visual similarity
- style
- transferring / A faster way to transfer style
- style matrix / Transferring style from one image to another
- supervised learning
- requisites / Learning algorithms
- about / Preparing the data
- supervision / What is machine learning?
- supported learning algorithms / Supported learning algorithms
- support vector machines (SVM) / Shadow draw – real-time user guidance for freehand drawing
- Swift
- custom layers, building / Building custom layers in Swift
- symbolic AI / What is machine learning?
T
- TensorFlowLite / Closing thoughts
- tokenization / iOS keyboard prediction – next letter prediction
- training
- about / What is machine learning?
- versus inference / Difference between training and inference
- transfer learning / Preparing the data
- Turi create
- reference / Closing thoughts
- typical workflow / A typical workflow
U
- U-Net architecture
- about / Classifying pixels
- reference / Classifying pixels
- unsupervised learning / Netflix – making recommendations
- use cases, machine learning
- speech recognition / Inference on the edge
- image recognition / Inference on the edge
- object localization / Inference on the edge
- optical character recognition / Inference on the edge
- translation / Inference on the edge
- gesture recognition / Inference on the edge
- text prediction / Inference on the edge
- text classification / Inference on the edge
- user's sketch
- recognizing / Recognizing the user's sketch
- training data, reviewing / Reviewing the training data and model
- model, reviewing / Reviewing the training data and model
V
- vector Digital Signal Processing (vDSP ) / Sorting by visual similarity
- vectorization / Accelerating our layers
- Vision / A brief introduction to Core ML
- visual bag of words / Shadow draw – real-time user guidance for freehand drawing
- Visual Object Classes (VOC) / Converting Keras Tiny YOLO to Core ML
- VNRecognizedObjectObservation
- reference / Making it easier to find photos
W
- woodblock printing / Transferring style from one image to another