[MLS-C01] [Algorithms] Image Analysis Algorithms

Defined

Supervised learning algorithms
Take images as input and either labels the images or identifies objects in the image
Two image analysis algorithms in the SageMaker built-in algorithms

Supervised learning algorithm that supports multi-label classification
Takes an image as input and outputs one or more labels assigned to that image
Uses a convolutional neural network (ResNet) that can be trained from scratch or trained using transfer learning when a large number of training images are not available
Can also seed the training of a new model with the artifacts from a model that you trained previously, called incremental training
Recommended input format is RecordIO, but can also use raw images in .jpg or .png format.
Example use case: classify image of person as on building entry security list
Important Hyperparameters
1. num_classes: number of output classes
2. num_training_samples: number of training examples in the input dataset
3. early_stopping: define a threshold at which to stop training

Supervised learning algorithm
Detects and classifies objects in images using a deep neural network
Takes images as input and identifies all instances of objects within the image
Object is categorized into one of the classes in a collection you specify, with a confidence score assigned to the class
Location and scale of the object in the image are noted by bounding box
Can be trained from scratch, or trained with models that have been pre-trained on the ImageNet dataset
Use input format of RecordIO, but can also use raw images in .jpg or .png format.
Example use case: detect objects in baggage at airport scan
Important Hyperparameters
1. num_classes: number of output classes
2. num_training_samples: number of training examples in the input dataset
3. use_pretrained_model: use a pre-trained model for training

本文由 Oscaner 创作, 采用知识共享署名4.0 国际许可协议进行许可
本站文章除注明转载/出处外, 均为本站原创或翻译, 转载前请务必署名