Lecanyu

Results 4 comments of Lecanyu

Image Format and Prep Instruction

I think the examples in the repository expose all required format. Please try to run and read those demos.

alpha.txt is missing

Please download the pretrained model [here](https://drive.google.com/open?id=1sUIcAzFTJNAAEEhqdYAKMKgzjVwRvsP4) which contains alpha.txt.

How do you create attention map?

Hi, have you found the solution about how to create attention map? I'm interested in this technique.

关于对位姿求导的一个简单问题

我疑惑的是为什么ceres这里的[ba优化例子](https://ceres-solver.googlesource.com/ceres-solver/+/master/examples/simple_bundle_adjuster.cc)（G2O里也有一个类似叫BAL的例子）9参数相机变量更新为什么可以直接用向量加法？ 因为里面的参数包含旋转向量，旋转向量v1+旋转向量v2 并不会准确的等于两个旋转的叠加（按李代数来看会差一个雅可比），这里难道是近似把雅可比认为是单位阵？