Metrics for Object Detection Introduction Summarize some most used metrics for object detection. Concept Confidence score is the probability that an anchor box contains an object. It is usually predicted by a classifier. Intersection over Union (IoU) is defined as the area of the intersection divided by the area of Read more…
1. FoundationIn feature point-based visual SLAM, we generally optimize the camera pose and map point location by minimizing the reprojection error. The following is an introduction to this issue.1.1 Map parameteriseTwo method: 3-d vector $X=\left[ x,y,z \right]^{T}$ , or inverse depth parametrization. Here we use 3-d vector for simplicity.1.2 Camera Read more…
Summary Lie algebra is useful in the SLAM problem. If the results of the map and pose are to be more accurate, an optimization problem is usually involved, and this optimization problem is often related to the solid body rotation or transformation matrix. But $SO(3)$ and $SE(3)$ do not satisfy Read more…
There are 3 categories of the most used methods to parameterize 3D solid body motion. i.e. matrix, quaternion and Lie group. Matrix Rotation $$ R=\begin{bmatrix} r_{11} &r_{12} &r_{13} \\ r_{21} &r_{22} &r_{23} \\ r_{31} &r_{32} &r_{33} \end{bmatrix}\in \mathbb{R^{3\times3}}, s.t. R~R^T=I, det(R)=1 $$ Transformation $$ T=\begin{bmatrix} R&t \\ 0^T&1 \end{bmatrix}\in \mathbb{R^{4\times4}} Read more…