Fix ts_ns unit bug in PointCloud LAS/LAZ loader - #473
Closed
michalpelka wants to merge 7 commits into
Closed
Conversation
We compile automatic jacoians (24 k chars headers) without debug symbols.
pair_wise_iterative_closest_point.cpp stayed in CORE_BASE_SOURCES while pose_graph_loop_closure.cpp (which calls PairWiseICP::compute) moved to CORE_MATH_SOURCES, splitting a symbol and its only definition across two static libs with a link-order dependency in the wrong direction. This broke linking for any executable that pulls in PoseGraphLoopClosure without referencing PairWiseICP directly, e.g. multi_view_tls_registration_step_2. pair_wise_iterative_closest_point.cpp also includes an auto-generated Jacobian header, so it belongs in core_math anyway. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DiH2pr8ruiHu6k7Y2wSXS2
pair_wise_iterative_closest_point.cpp (moved into core_math in the previous commit) calls get_rgd_index_3d(), which lived in hash_utils.cpp under CORE_BASE_SOURCES -- reintroducing the same cross-archive circular dependency, just in the opposite direction (core_math needing a symbol from core/core_no_gui instead of the other way around). Nothing else in CORE_BASE_SOURCES/CORE_GUI_SOURCES calls into core_math, so moving hash_utils.cpp there too makes the dependency one-directional again (core/core_no_gui -> core_math, never the reverse). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DiH2pr8ruiHu6k7Y2wSXS2
…lors Ports the three raylib/ImGui tools from the sibling mandeye-colors repo as new apps (camera_lidar_calibration, camera_lidar_trajectory_viewer, camera_lidar_intrinsics_calib), backed by a new calib_core static library for their shared non-GUI logic (camera projection math, LAS/LAZ loading, trajectory CSV parsing, CLI args). Reuses HDMapping's existing raylib/ imgui_raylib/rlimgui/Eigen/LASzip/OpenCV/json wiring instead of vendoring mandeye-colors' own duplicate copies of those dependencies. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DiH2pr8ruiHu6k7Y2wSXS2
All three text fields that take a file or directory path (image, point cloud, intrinsics, calibration, session/CAMERA_0 dirs, export/ROS/COLMAP output paths) now have a "Browse..." button backed by portable-file-dialogs, matching the mandeye::fd wrapper core already uses elsewhere in HDMapping. Implemented as calib_core's own calib::fd (rather than reusing core/include/Core/pfd_wrapper.hpp directly) since that wrapper only builds into the GUI-enabled `core` target, and linking `core` here would pull in core_math/session/SLAM code these apps otherwise don't depend on -- portable-file-dialogs itself is a single vendored header with no dependency on `core`, so wrapping it directly in calib_core is cheap. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DiH2pr8ruiHu6k7Y2wSXS2
… apps Re-ports camera_lidar_trajectory_viewer's TrajectoryViewer.cpp from mandeye-colors' mp/roi branch (the initial import was mistakenly based on master, before the ROI work landed there): adds a Roi struct (calib_core's Camera.h), SLERP pose interpolation instead of nearest-neighbor lookups, distortion-aware point projection, a "Geometry" coloring strategy (closest camera by depth) alongside the existing temporal one, ROI-filtered colorization with a Camera-ID/In-ROI render mode and an ROI overlay on the image preview, and colored/uncolored point-count stats. Also fixes three real bugs surfaced by manually running the apps: - Fixed-size windows (1400x900 etc.) could be taller than the screen once the OS menu bar + title bar are accounted for, silently pushing the top of the control panel off-screen. Added fitWindowToScreen() (monitor-aware resize/reposition) and SetWindowMinSize() to all three apps. - fitWindowToScreen() itself was buggy: GetMonitorWidth/Height return the monitor's native pixel resolution while GetScreenWidth/Height and SetWindowSize/SetWindowPosition operate in logical points, a 2x mismatch on Retina displays that pushed the window mostly off-screen. Fixed by dividing by GetWindowScaleDPI(). Also added a PollInputEvents() call so raylib's cached mouse/window geometry is refreshed before rlImGuiSetup() reads it. - Several ImGui widgets with trailing labels (InputInt, Combo, InputDouble) were wrapped in PushItemWidth(-1), which gives the widget box the entire row and clips the label off the right edge of the panel. Scoped a narrower width around each affected widget. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DiH2pr8ruiHu6k7Y2wSXS2
gps_time from the LAS/LAZ point record is in seconds, but Point3D::ts_ns is treated as nanoseconds everywhere it's consumed (Trajectory::nearest(), TrajectoryViewer's gps_time round-trip on export, nearest-image-by-timestamp matching). Without the *1e9 conversion, loaded point timestamps were off by ~9 orders of magnitude from every value they get compared against, making nearest-image/nearest-pose matching against point timestamps meaningless. Fixes MapsHD#472 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DiH2pr8ruiHu6k7Y2wSXS2
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
calib::PointCloud::load()castgps_time(seconds, per LAS spec) directly tots_nswithout converting to nanosecondsts_nsis treated as nanoseconds everywhere else it's used:Trajectory::nearest(), the gps_time round-trip on export inTrajectoryViewer.cpp, and nearest-image-by-timestamp matching* 1e9conversion, loaded point timestamps were off by ~9 orders of magnitude from every value they get compared against, making nearest-image/nearest-pose matching against point timestamps meaninglessFixes #472
Test plan
.lazfile throughcamera_lidar_calibrationorcamera_lidar_trajectory_viewerand confirm a loaded point'sts_nsis now in the same order of magnitude as the corresponding trajectory/image timestamp🤖 Generated with Claude Code
https://claude.ai/code/session_01DiH2pr8ruiHu6k7Y2wSXS2