From 464845a19f9d7e528f55e966022743d29d0cc976 Mon Sep 17 00:00:00 2001 From: "Yi-Hsiang Fang (Vivid)" <146902905+vividf@users.noreply.github.com> Date: Fri, 3 May 2024 11:11:57 +0900 Subject: [PATCH] Update sensor/docs/tutorials/tag_based_pnp_calibrator.md Co-authored-by: Kenzo Lobos Tsunekawa --- sensor/docs/tutorials/tag_based_pnp_calibrator.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/sensor/docs/tutorials/tag_based_pnp_calibrator.md b/sensor/docs/tutorials/tag_based_pnp_calibrator.md index f50a465a..b04f931f 100644 --- a/sensor/docs/tutorials/tag_based_pnp_calibrator.md +++ b/sensor/docs/tutorials/tag_based_pnp_calibrator.md @@ -142,7 +142,7 @@ The images below show that with the calibrated transformation, the projected poi - Why does the reprojection error increase when more data is collected? - 1. When there are few samples, the model will fit the available data the best it can, even in the presence of noise (over-fitting). The more data is collected, the error may increase to a certain extent, but that is the model trying to fit all the data, being unable to fit the noise. However, it should reach a more-or-less table peak with about 10-15 pairs (depending on the data collection pattern/sampling). + 1. When there are few samples, the model will fit the available data the best it can, even in the presence of noise (over-fitting). The more data is collected, the error may increase to a certain extent, but that corresponds to the model attempting to fit all the data, this time unable to fit the noise, resulting in a higher error. However, it should reach a more-or-less table peak with about 10-15 pairs (depending on the data collection pattern/sampling). - Why doesn't the reprojection error seem low enough?