- base_connaissances/ → snippets/ (corpus/ → textes/, instructions.md → consignes_redaction.md, decoupage.py → fichiers.py, importer.py → envoyer.py) ; python3 -m snippets.envoyer. - instance-de-test/ : 1_demarrer.sh, 2_creer_boite_et_cle.py, 3_brancher_ia.py, 4_essayer_brouillons.py, voir_empreintes.sh, connexion_admin.py. - ollama/verifier.py → ollama/verifier_ollama.py ; tests, tâches VS Code et docs alignés. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
19 lines
893 B
Markdown
19 lines
893 B
Markdown
# How should reference targets be selected?
|
|
|
|
Where possible, use several reference targets distributed over the image rather than relying on a
|
|
single target.
|
|
A distributed reference network makes camera-motion estimation more robust and provides
|
|
redundancy if an individual target becomes obstructed, damaged, or moves.
|
|
Reference targets should preferably:
|
|
• be clearly visible;
|
|
• provide good image contrast;
|
|
• remain visible under the expected illumination conditions;
|
|
• be distributed over the monitored image;
|
|
• be sufficiently far from the camera where appropriate;
|
|
• not all belong to a small local area that could deform as a single block.
|
|
When no completely stable point is available, a larger reference network can still be used
|
|
because camera-motion estimation is based on a robust fit and its consistency can be monitored
|
|
using sigma0.
|
|
|
|
Source : AIMsight FAQ, rév. 1.5
|