Guibin Chen

代码中没有用Levenberg-Marquardt算法(L-M method)，用的是高斯牛顿法，虽然paper中说用的LM

If you meet this problem, please exe the follows: git reset --hard 338852b4f67e32d6c3dadfdedc6cc40ca7eef7f6

the accuracy of pose in kitti training data

4

the translation error can reach 0.88% as described in Zhang Ji ’s paper？