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

This project consist from:

dope4j - Java library which can be used to run inference with DOPE models and detect object poses.

dope4j-app - Java application which exposes dope4j library functionality through CLI.

Download

Release versions

Or you can add dependency to it as follows:

Gradle:

dependencies {
  implementation 'io.github.lambdaprime:dope4j:2.0'
}

Documentation

Documentation

Development

Usage

dope4j-app -action=<runInference|showResults> <options>

runInference

Options:

-modelUrl=<path>
-objectSize=<width>,<height>,<length>
-imagePath=<path>
-cameraInfo=<path>
-showVerticesBeliefs=<true|false>
-showCenterPointBeliefs=<true|false>
-showAffinityFields=<true|false>
-showMatchedVertices=<true|false>
-showCuboids2D=<true|false>
-showProjectedCuboids2D=<true|false>
-cache=<true|false>
-cacheFolder=<path>
-recursiveScan=<true|false>
-imageFileRegexp=<regexp>
-threshold=<double>
-debug=<true|false>
-exportMetricsToCsv=<path>
-exportMetricsToElastic=<elasticsearch_url>
-totalRunTime=<true|false>
-lineThickness=<int>

showResults

Options:

-resultsJson=<path>
-imagesRoot=<path>

Contributors

lambdaprime [email protected]