dope4j - provides access to Deep Object Pose Estimation (DOPE) models from Java. Based on Deep Java Library.

Usage

// cuboid model of ChocolatePudding which network was trained on
var objectCuboidModel = new Cuboid3D(new Point3D(),
  4.947199821472168, 2.9923000335693359, 8.3498001098632812);
var dopeService = new Dope4jFactory().createPoseEstimationService(
  "/models/ChocolatePudding.onnx", objectCuboidModel, cameraInfo);
System.out.println(dopeService.analyze(imageFile));

To create "ChocolatePudding.onnx" follow the instructions from Quickstart to isaac_ros_pose_estimation.

Metrics

dope4j integrated with OpenTelemetry for emitting metrics. OpenTelemetry is similar to SLF4J in that it serves as a simple facade or abstraction for various observability backends (Prometheus, Elasticsearch etc). On top of that it is available for many programming languages.

If you plan to export metrics to Elasticsearch or CSV files you can try opentelemetry-exporters-pack. For complete list of available exporters for OpenTelemetry see Registry for OpenTelemetry ecosystem.

Documentation
Download
GitHub repository

Packages

Package

Description

id.deeplearningutils.modality.cv.output

Additions of primitive types to Deep Java Library

Main package

dope4j 2.0 API

Usage

Metrics