ai.djl.mxnet.zoo.cv.objectdetection

Class SingleShotDetectionModelLoader

java.lang.Object

ai.djl.repository.zoo.BaseModelLoader<ai.djl.modality.cv.Image,ai.djl.modality.cv.output.DetectedObjects>

ai.djl.mxnet.zoo.cv.objectdetection.SingleShotDetectionModelLoader

All Implemented Interfaces:

ai.djl.repository.zoo.ModelLoader<ai.djl.modality.cv.Image,ai.djl.modality.cv.output.DetectedObjects>

public class SingleShotDetectionModelLoader
extends ai.djl.repository.zoo.BaseModelLoader<ai.djl.modality.cv.Image,ai.djl.modality.cv.output.DetectedObjects>

A ModelLoader for Single Shot Detection (SSD) models.

These models were built as part of the Gluon CV library and imported into DJL.

SSD is a model to solve Application.CV#OBJECT_DETECTION. Prior models before SSD typically used non-deep learning strategies to find object proposals and then would execute classification at those proposed locations. SSD integrated the proposal process into the model and then classifies at all of those locations in parallel. This both simplifies the model and training process while better leveraging the power of model DL engines. [paper]

Today, SSD is not typically used in favor of newer models which outperform it.

Field Summary

Fields inherited from class ai.djl.repository.zoo.BaseModelLoader

factories, modelZoo, mrl, repository, version

Constructor Summary

Constructors

Constructor and Description
SingleShotDetectionModelLoader(ai.djl.repository.Repository repository) Creates the Model loader from the given repository.

Method Summary

Modifier and Type	Method and Description
ai.djl.Application	getApplication()
ai.djl.repository.zoo.ZooModel<ai.djl.modality.cv.Image,ai.djl.modality.cv.output.DetectedObjects>	loadModel(java.util.Map<java.lang.String,java.lang.String> filters, ai.djl.Device device, ai.djl.util.Progress progress) Loads the model with the given search filters.

Methods inherited from class ai.djl.repository.zoo.BaseModelLoader

createModel, getArtifactId, listModels, loadModel, match, toString

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Methods inherited from interface ai.djl.repository.zoo.ModelLoader

loadModel, loadModel, loadModel, loadModel

Constructor Detail

SingleShotDetectionModelLoader

public SingleShotDetectionModelLoader(ai.djl.repository.Repository repository)

Creates the Model loader from the given repository.

Parameters:

repository - the repository to load the model from

Method Detail

getApplication

public ai.djl.Application getApplication()

loadModel

public ai.djl.repository.zoo.ZooModel<ai.djl.modality.cv.Image,ai.djl.modality.cv.output.DetectedObjects> loadModel(java.util.Map<java.lang.String,java.lang.String> filters,
                                                                                                                    ai.djl.Device device,
                                                                                                                    ai.djl.util.Progress progress)
                                                                                                             throws java.io.IOException,
                                                                                                                    ai.djl.repository.zoo.ModelNotFoundException,
                                                                                                                    ai.djl.MalformedModelException

Loads the model with the given search filters.

Parameters:

filters - the search filters to match against the loaded model

device - the device the loaded model should use

progress - the progress tracker to update while loading the model

Returns:

the loaded model

Throws:

java.io.IOException - for various exceptions loading data from the repository

ai.djl.repository.zoo.ModelNotFoundException - if no model with the specified criteria is found

ai.djl.MalformedModelException - if the model data is malformed