[simheo]

Stacked on #26: the base is postprocess-consolidation, so the diff shows only this change. When #26 merges, change the base to develop (GitHub does it automatically if #26's branch is deleted on merge).

The Orbbec Gemini 305 is the default depth camera (depth_camera = "gemini305" in grabette's config), but the post-processing still described every recording as an OAK-D's. This PR is the prose-only step of the OAK → generic cleanup.

What changes

• READMEs (root + grabette-postprocess/): the intro, data flow and SLAM description now say "depth camera" and spell out what depends on an IMU (visual-inertial with the OAK-D SR, visual-only with the Gemini 305). The Hardware section lists both cameras plus the RPi fisheye.

• Docstrings / comments in checks/sync.py, checks/recording.py, checks/__init__.py, convert.py, pipeline.py and the CLI scripts.

• User-visible strings:

  • the Space's sync message OAK cam↔gyro lag → depth-cam ↔ gyro lag
  • the check labels OAK-left ↔ OAK-gyro → depth-cam left ↔ gyro
  • the recording-check messages oak imu: … → imu: …, matching the existing imu: none line

  No code or test matches on any of these strings.

• Env var: OAK_VSLAM_BINARY → VSLAM_BINARY. It's set in the Dockerfile and read in pipeline.py, renamed together, so the Space behaves the same.

Deliberately not in this PR

• Identifiers, files and directories (oak_slam.py, run_oak_slam(), check_oak_imu(), convert_episode_to_oak.py, the oak/ intermediate dir, docker/oak_vslam/, the oak_dir variables): these wait for feat/tactile-sensing, casquette: raw camera-only pass-through in the multi-arm build #5 and 22 modify the threshold for trajectory check #23 to land, so they don't hand those branches a painful rebase.
• The Docker image pollenrobotics/oak-vslam → pollenrobotics/grabette-vslam: needs a Docker Hub push first, and goes with the code renames.
• Lines inside in-flight hunks are left alone: app.py:84 and pipeline.py:433, plus everything in dataset.py, oak_slam.py, trajectory.py, checks/tags.py and checks/trajectory.py.
• Text that really is about the OAK-D: the legacy oakd_* file names and their tests (old recordings still use them), the IMU/OAK-D-specific notes in checks/recording.py, the measurement notes in compare_trajectories.py, the DepthAI attribution, and the "oak" RTAB-Map camera label in offline_vslam.cpp.

Verification

• pytest: 67 passed
• Every changed .py file compiles. --help runs for check_sync.py, check_dataset.py, visualize_rgbd_trajectory.py (with --extra viz) and run_oak_slam.py.
• No file touched by feat/tactile-sensing, casquette: raw camera-only pass-through in the multi-arm build #5 or 22 modify the threshold for trajectory check #23 in the library, app.py or views.py. The edits in pipeline.py and Dockerfile are outside their hunks.
• Test Space build: not run. The Dockerfile change is only the ENV name, read by pipeline.py in the same commit.

🤖 Generated with Claude Code