Currently we align different perspectives with their bounding boxes. This means that outliers can lead to misalignment of measurements. A better approach might be to use linear regression to pick the best scale and shift to merge two perspectives. The important thing is to only use the points with high confidence in both perspectives for the linear regression.
Currently we align different perspectives with their bounding boxes. This means that outliers can lead to misalignment of measurements. A better approach might be to use linear regression to pick the best scale and shift to merge two perspectives. The important thing is to only use the points with high confidence in both perspectives for the linear regression.