Sameer Agarwal

This looks like an uninitialized summary object. If your optimization succeeded the. You should not get a summary like this.

so there are two solves being called? so this has nothing to do with the first successful solve. Please print Summary::FullReport it will tell you what is going on. also...

cc: @joydeep-b

Yes of course. Please send us a pull request based on the instructions on ceres-solver.org. On Thu, Jun 18, 2020, 3:30 PM Ethan Smith wrote: > Hi! I'd like to...

Thanks for the patch, could you submit to gerrit using the instructions at http://ceres-solver.org/contributing.html

This has nothing to do with number of images. You have a cost function which is being fed values 0.996067, -0.0300342, -0.0493301 , -0.0671961 -20.0864, -5.85186, -7.55326 19.814, 8.13496, 5.99895...

I suggest evaluating the cost function on its own and add logging to the cost function to print out the intermediate values. when automatic differentiation is used, logging the values...

Even with @alexsmac's fix to the line search #6851a63 this still remains a problem.

@dellaert you are right and this is a wart in the ceres API, that I haven't quite figured out how to fix in a proper fashion. Manifold support came to...

Try this, in residual_block.cc replace ``` // Update the jacobians with the local parameterizations. if (jacobians != nullptr) { for (int i = 0; i < num_parameter_blocks; ++i) { if...