Krishna Murthy

Are you running with autograd enabled? In which case gradslam will try to maintain a computation graph and you'll eventually run out of memory. You can get around this a...

Thanks @Yarp96 ! The step-by-step Pointfusion approach is useful when you're using gradslam in an interactive application (e.g., a robot navigating around a room captures an image, decides on a...

Thanks for your interest @mohdomama One aspect the gradslam release can benefit from is further documentation and examples on leveraging the gradients. There are a few repos that already document...

@MisEty all good questions! 1. The reason we use LM (as opposed to solving the original normal equations) is because the approximate Hessian `A = J^T J` can often turn...

1. `A` is the Jacobian -- the Hessian would infact be the matrix of second derivatives, but since that's too cumbersome to compute, people approximate the hessian by `A^T A`....

Thanks for the question @adizhol Currently there isn't an easy way to do this, but I can envisage a few use-cases for this. Could be a good feature to add

Hi @apoorvkh, Indeed, processing speeds and code optimization are some of our focus areas in subsequent releases. (and yes, decreasing `numiters` is expected to cause degradation in reconstruction quality) One...

@SilvioGiancola KinectFusion is indeed on our roadmap for version 0.2.0

We're actively working on it -- although we want to get a sense of a stable API and feature requests first My estimate would be first or second week of...

@TontonTremblay I've reimplemented in PyTorch: https://github.com/krrish94/nerf-pytorch