Martin de La Gorce

In the camera projection backward pass I don't compute derivatives w.r.t the cameras extrinsic parameters ( in [project_points_backward](https://github.com/martinResearch/DEODR/blob/8aa25cd442db080c7c50fde1534956ac9b0d6c68/deodr/differentiable_renderer.py#L302) I compute only derivatives w.r.t the points positions *points_3d_b*) so we can't...

If you have only one camera it is not possible to estimate the size of the object: you could have a small object close to the camera or a big...

In order to estimate the camera positions with a known object I would not use my library but opencv. Mesh fitting through differentiable rasterisation is not the right tool here....

Hi, DEODR does not support multiple meshes at the moment. This is in the "to do" list. In the mean time you will have to pose each mesh and combine...

For some reason matlab's kronSparse seems not to take into account the fact that the matrix will be sparse when computing the memory footprint of the resulting matrix. One option...

@d-cogswell , it would be great if you could provide me with some matrices A,B and vector v to replicate the problem. Did you try A*(v.*B)?

Using a different ordering of the multiplications, it runs fast for me: ``` f3 = @(x) A*(x.*(A*x)); tic J3=AutoDiffJacobianAutoDiff(f3, u); toc ``` it runs in 0.16 sec on my machine...

Ok. Maybe this behaviour could be controlled trough an new optional argument with default value set up to keep default behaviour unchanged to avoid breaking code that relies on it...

I see. Is that something you would want to do ?

It looks like it is now using `is_close` with` abs_tol` whenever math_epsilon is not none https://github.com/seperman/deepdiff/blob/7e5e9954d635423d691f37194f3d1d93d363a128/deepdiff/diff.py#L1116 However It would be great to add another parameter that would allow to use...