Parameter Tunning Guide

This guide will walk through some rules of thumb when it comes to adjusting parameters to get better detection.

Not detection all objects

Increase the search detail. This will increase the number of computed scene features
Decrease acceptance threshold. This lowers the acceptance threshold of the score filter, keeping poses with lower scores

../../_images/search_detail_low.png

../../_images/search_detail_high.png

../../_images/acceptance_low.png

../../_images/acceptance_high.png

False positive

Reduce downsample strength. This makes the downsample voxel size smaller, preserving more detail in the scene. Reducing downsample strength can also improve poses that are badly oriented.

../../_images/downsample_low.png

downsample strength low

../../_images/downsample_medium.png

downsample strength medium

Model parameters

Adjust model feature detail. Increasing model feature detail will increase the amount of model features calculated, making it more likely to accurately find the object in the scene
Adjust model downsample strength. Reducing downsample strength can remove false positives

../../_images/feature_detail_low.png

feature detail low

../../_images/feature_detail_high.png

feature detail high

../../_images/model_downsample_low.png

downsample strength low

../../_images/model_downsample_medium.png

downsample strength medium