All issues

Search results

RampNet 2.0 should consume the canonical cropper, not its own: Stage 1's crop supervision inherits the tilt error in Project Sidewalk's pano_y

Status: Open.

#113 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Aug 6, 2026

The recall-by-distance axis is stretched ~25-30%: recompute it from GSV depth

Status: Open.

#112 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Aug 5, 2026

GSV serves a native per-pano depth map (streetlevel already parses it): an independent anchor for #101 and the visibility test #102 needs — archive it before the endpoint changes

Status: Open.

#111 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Aug 5, 2026

Swap the Stage 2 backbone to DINOv3-ConvNeXt (there is no ConvNeXt V3) - size-matched pretraining comparison, licence-gated

enhancementNew feature or requestNew feature or request

Status: Open.

#107 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Aug 4, 2026

Review location precision against the STREET-LEVEL imagery Stage 1 actually consumes, not an aerial proxy (#96)

documentationImprovements or additions to documentationImprovements or additions to documentation

Status: Open.

#103 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Aug 3, 2026

Mine hard positives from multi-view consensus: a training-label source that does not depend on city inventories (#96, #59)

Status: Open.

#102 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Aug 3, 2026

Flat-ground distance carries a ~11% range-proportional scale error, not just scatter — multi-view self-consistency measures it and DA3 can anchor it

Status: Open.

#101 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Aug 3, 2026

Assess location precision for the unassessed candidate cities — the gate every route to 500k depends on

documentationImprovements or additions to documentationImprovements or additions to documentation

Status: Open.

#96 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Jul 31, 2026

Attribute any RampNet-vs-YOLO gap: backbone, output representation, and training budget are confounded in #51

Status: Open.

#95 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Jul 31, 2026

Tune the YOLO baseline's peak LR / warmup before reporting it — the collapse is an untuned default (#51)

Status: Open.

#90 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Jul 30, 2026

RampNet 2.0: measurement, condition inference, and tagging — we already collect the supervision and discard it at ingest

Status: Open.

#86 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Jul 29, 2026

Aerial orthoimagery curb-ramp detection already exists in production (34k ramps at 1-inch GSD): treat it as an uncorrelated arbiter, not a competitor

Status: Open.

#85 In ProjectSidewalk/RampNet;

· jonfroehlich opened on Jul 29, 2026