Hypochondria
Hypochondria
勾选了MNN_CUDA,MNN_USE_SYSTEM_LIB,取消了MNN_SEP_BUILD,ubuntu18上成功编译,推理时采用cuda后端还是报Can't Find type=2 backend, use 0 instead
https://github.com/TimoBolkart/BFM_to_FLAME/blob/5d12b6ba9d4c4dfed387e26a9b25780dbd00adf6/mesh_convert.py#L74 Can this parameter convert the units of BFM into meters? Because I noticed FLAME is in Meters.
how to get true yaw, pitch, roll annotation from current dataset?
Hi,I am studying 3d face reconstruction based 3DMM and I am curious about the objective function and the construction of Jacobian matrix.Do you have some references? Thank you!
情况是这样，如果GT模型h和预测模型p分辨率差很多，比如GT模型一般可以用扫描仪获得，顶点数远远大于预测的模板模型。看论文bICP应该查找的也是一对一的点对，也就是最终计算的点对数是预测模板p的个数，这种评价方法很正确。我想请问，如果这个时候要用gICP，计算p->h方向比h->p方向应该更能反应真实的误差吧？我理解h->p方向GT模型h上会有很多点找到了预测模型p上面的同一点。
Hello, this is a great project. Can you provide more detailed documentation?
In your code, I found that the nodes are evenly sampled by the PCL library，but I didn't figure out how the connection between the nodes was determined.
Looking forward to the possibility of open-sourcing.