after "solver.graph.optimize()"

[huzhongyyuan]

How can I save the predictions after performing SL(4) alignment using the solver.run_predictions method? I need the complete and continuous parameters for each frame (3000 frames in total) after they have been aligned and optimized. Specifically, how can I access and save the predictions after calling solver.graph.optimize()

[Dominic101]

I don't understand the question. What are the predictions and parameters you are referencing?

[huzhongyyuan]

I am working with submaps, where each sub_map corresponds to the predictions from solver.run_predictions(image_names_subset, model, args.max_loops). From each sub_map, I can obtain camera parameters and depth maps. However, I’m wondering if the camera parameters across different submaps are consistent or if they need additional alignment. If I want to achieve a continuous camera trajectory across submaps, should I apply any specific post-processing steps to ensure proper alignment and continuity of the camera trajectory? Would you be able to clarify if my understanding is correct, and if any further adjustments are necessary?

I have noticed that the matching between submaps seems to have primarily focused on point clouds, without explicitly handling the alignment of camera parameters. To achieve a continuous camera trajectory that covers the entire set of submaps, I would like to know the appropriate method for obtaining the aligned camera positions, extrinsic and intrinsic parameters, and depth maps across the submaps. Could you provide guidance on how to ensure that these parameters are consistently aligned across submaps for a continuous and accurate trajectory?

Much Thanks！！！