| name | photo-to-replan |
| description | Use this skill when the user wants to run the end-to-end all-sky photo pipeline from a raw all-sky image to a replanned observing sequence, using the packaged run_pipeline.py entrypoint in this shared directory. |
Photo To Replan
Use this skill when the task is to go from a raw all-sky image to replanned observing outputs.
Workflow
-
Locate the shared photo_to_replan_pipeline directory.
-
Run the single-entry pipeline script from that directory:
python run_pipeline.py --image /abs/path/to/image.jpg
Optional:
- Override parameter group with
--position-name position_X
- Save segmentation overlay with
--save-overlay
- Read the generated report:
output/<image_stem>/pipeline_report.json
- From the report, inspect these outputs when needed:
scenario.jsonreplanned_schedule.jsondeferred_targets.jsonreplanned_sequence.ninaTargetSetsky_replan_plot.png
Input Requirements
- The input must be a raw all-sky camera photo.
- The filename should contain Beijing time, preferably in
YYYY_MM_DD_HH_MM_SS.jpg form.
Notes
- The mask stage uses the packaged
Pic2mask/full_image_infer/infer_full_image.py.
- The replan stage uses the packaged
sequence_adjust_tool/scheme_horizon_to_image/replan.py.
- The pipeline configuration is stored in
configs/default.json.
- The packaged configuration uses a single conda environment for the whole pipeline.
Validation
A run counts as pipeline-successful if these files exist and are readable:
pipeline_report.jsonscenario.jsonreplanned_schedule.jsonreplanned_sequence.ninaTargetSet
Business success is separate. Check deferred_targets.json and replanned_schedule.json to see whether future slots were actually filled.
