ai.djl.basicmodelzoo.cv.object_detection.ssd

Class SingleShotDetection.Builder

java.lang.Object

ai.djl.basicmodelzoo.cv.object_detection.ssd.SingleShotDetection.Builder

Enclosing class:

SingleShotDetection

public static class SingleShotDetection.Builder
extends java.lang.Object

The Builder to construct a SingleShotDetection.

Method Summary

Modifier and Type	Method and Description
SingleShotDetection	build() Builds a SingleShotDetection block.
SingleShotDetection.Builder	optFeatures(java.util.List<ai.djl.nn.Block> features) Sets the Conv2d blocks to be appended to the network to get multi-output network.
SingleShotDetection.Builder	optGlobalPool(boolean globalPool) Sets the boolean whether to attach a global average pooling layer as the last output layer.
SingleShotDetection.Builder	setBaseNetwork(ai.djl.nn.Block network) Sets the base network for the SSD framework.
SingleShotDetection.Builder	setNumClasses(int numClasses) Sets the number of classes of objects to be detected.
SingleShotDetection.Builder	setNumFeatures(int numFeatures) Sets the number of down sampling blocks to be applied.
SingleShotDetection.Builder	setRatios(java.util.List<java.util.List<java.lang.Float>> ratios) Sets the list of aspect ratios of generated anchor boxes.
SingleShotDetection.Builder	setSizes(java.util.List<java.util.List<java.lang.Float>> sizes) Sets the list of sizes of generated anchor boxes.

Methods inherited from class java.lang.Object

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

Method Detail

setSizes

public SingleShotDetection.Builder setSizes(java.util.List<java.util.List<java.lang.Float>> sizes)

Sets the list of sizes of generated anchor boxes.

Parameters:

sizes - size of the input

Returns:

Returns this Builder

setRatios

public SingleShotDetection.Builder setRatios(java.util.List<java.util.List<java.lang.Float>> ratios)

Sets the list of aspect ratios of generated anchor boxes.

Parameters:

ratios - size of the input

Returns:

Returns this Builder

setNumClasses

public SingleShotDetection.Builder setNumClasses(int numClasses)

Sets the number of classes of objects to be detected.

Parameters:

numClasses - number of classes

Returns:

Returns this Builder

setBaseNetwork

public SingleShotDetection.Builder setBaseNetwork(ai.djl.nn.Block network)

Sets the base network for the SSD framework.

Parameters:

network - Base network

Returns:

Returns this Builder

setNumFeatures

public SingleShotDetection.Builder setNumFeatures(int numFeatures)

Sets the number of down sampling blocks to be applied. Down-sampling blocks are applied to the base network successively, and feature maps are drawn from the each of the blocks. This value is ignored if features is also set.

Parameters:

numFeatures - Number of down sampling blocks to be applied

Returns:

Returns this Builder

optFeatures

public SingleShotDetection.Builder optFeatures(java.util.List<ai.djl.nn.Block> features)

Sets the Conv2d blocks to be appended to the network to get multi-output network.

Parameters:

features - List of Conv2d blocks to be appended

Returns:

Returns this Builder

optGlobalPool

public SingleShotDetection.Builder optGlobalPool(boolean globalPool)

Sets the boolean whether to attach a global average pooling layer as the last output layer.

Parameters:

globalPool - Whether to attach a global average pooling layer as the last output layer

Returns:

Returns this Builder

build

public SingleShotDetection build()

Builds a SingleShotDetection block.

Returns:

the SingleShotDetection block