Noah
Noah
> What version of GCC are you using? Hi rserban, I'm using 7.4.0
Paper title: Lite-Mono: A Lightweight CNN and Transformer Architecture for Self-Supervised Monocular Depth Estimation Paper link: https://arxiv.org/abs/2211.13202 Code link: https://github.com/noahzn/Lite-Mono
@acai66 Hi, thank you for your contribution. I found that only `detectAndComputeDense` is supported. Could you also make it support for `detectAndCompute`?
@acai66 Thank you very much!
@acai66 Have you met a problem in `DetectAndCompute` that the `#Select top-k features` function exceeded the max number when using TRT Execution Provider.
@Phil26AT I have met the same problem. Increasing the early stopping confidence doesn't help.
@guipotje Have you tried training Xfeat and LightGlue end-to-end?
> Hello @noahzn, I haven't tried to train it end-to-end. It might deliver some improvements, as mentioned in SuperGlue's paper (section 5.4), when backpropagating through the descriptors. However, it also...
Hi, @georg-bn thanks for the work! I am wondering how I should use a learned steerer pth file and use the descriptors to train lightglue, instead of using MNN?
@georg-bn Hi, thank you for your reply! Yes, I'm interested in this project. Now I'm training XFeat with a learned steerer.