EfficientDetD3

This model is a real-time neural network for object detection that detects 90 different classes (labels are available in org.jetbrains.kotlinx.dl.dataset.handler.cocoCategories).

Internally it uses the EfficientNets as backbone networks.

The model have an input with the shape is (1x896x896x3) by default. H and W could be changed by user to any values.

The model has 1 output:

  • detections:0 with 7 numbers as [unknown number, ymin, _xmin_, ymax, xmax, score, coco label].

NOTE: The detections are limited to 100.

See also

Functions

model
Link copied to clipboard
open fun model(modelHub: ModelHub): OnnxInferenceModel
preInit
Link copied to clipboard
open override fun preInit(): EfficientDetObjectDetectionModel
preprocessInput
Link copied to clipboard
open fun preprocessInput(imageFile: File, preprocessing: Preprocessing): FloatArray
open override fun preprocessInput(data: FloatArray, tensorShape: LongArray): FloatArray
pretrainedModel
Link copied to clipboard
open override fun pretrainedModel(modelHub: ModelHub): EfficientDetObjectDetectionModel

Properties

channelsFirst
Link copied to clipboard
open override val channelsFirst: Boolean
modelRelativePath
Link copied to clipboard
open override val modelRelativePath: String