Gate Part 1: Merging Exisiting XFeat Orientation Repo + Reorganizing Gate to Propsed Structure

[harshi-puli]

Summary

Reorganizes the gate task to match the classical//yolo//tests/ convention and merges in the gate-orientation prototypes from raymondt31/UR-B-Perception.

• Reorganize classical gate algorithms: moved GateCenterAlgo.py, GateSegmentationAlgoA/B/C.py, and the old archive/ scripts into gate/classical/. Added empty gate/yolo/ and gate/tests/ as placeholders (no YOLO model or tests exist yet). Updated perception/__init__.py's ALGOS imports to the new paths.
• Port gate orientation + XFeat runtime: added gate/orientation/, ported as-is from UR-B-Perception:
  • classical_orientation.py — LAB-color segmentation approach for post-detection orientation refinement (takes a YOLO box as input).
  • xfeat_orientation.py — XFeat-based homography matching against a static reference image, live webcam loop.
  • modules/ — the XFeat model itself (xfeat.py, model.py, interpolator.py, lighterglue.py), adapted for package-relative imports.
• Dependency: added tqdm to requirements-torch.txt (XFeat dependency).

Notes

• Mechanical port, not a rewrite — orientation code isn't yet wrapped to the TaskPerceiver interface or tested; tracked as follow-up.
• No behavior change to the existing classical gate algorithms, only their location.

Test plan

• pytest passes (classical gate algorithms unaffected)
• ALGOS imports resolve (gateseg, gatesegA/B/C)
• Manually verify classical_orientation.py/xfeat_orientation.py still run standalone post-move