Runtime and resources

Timing first, resource next

The SOP requires tracking, estimation, and total timing now. CPU, memory, and GPU profiling are the staged resource track and stay TBD until monitored reproducibly.

Public artifacts

SOP-aligned profiling outputs

This page uses the SOP timing format and separates required timing from next-stage CPU, memory, and GPU monitoring.

Required timing output

timing/<system>/<dataset>/<sequence>.txt

# timestamp(s),tracking,estimation,total

Required by the SOP after timers are added to visual tracking and estimation code paths.

Pose output paired with timing

algorithms/<system>/<dataset>/<sequence>.txt

# timestamp(s) tx ty tz qx qy qz qw

Timing is not a replacement for trajectory evaluation; both files must point to the same run.

Next-stage resource log

resource/<system>/<dataset>/<sequence>/bm_logger.csv

timestamp(s),cpu_usage,memory_usage,gpu_usage_if_applicable

CPU, memory, and GPU remain TBD until the BMLogger-style monitor and platform metadata are released.

Platform note

resource/<system>/<dataset>/<sequence>/platform.json

cpu, memory, gpu, operating_system, docker_image

Resource rows are not comparable without the platform note and container identity.

File schema

TBD-safe profiling templates

Templates define the column contract before numeric profiling results are released.

timing/<system>/<dataset>/<sequence>.txt

# timestamp(s),tracking,estimation,total
TBD,TBD,TBD,TBD

resource/<system>/<dataset>/<sequence>/bm_logger.csv

# timestamp(s),cpu_usage,memory_usage,gpu_usage_if_applicable
TBD,TBD,TBD,TBD