From a3e08a9bbcbe574fbb5736de87898766d6704965 Mon Sep 17 00:00:00 2001 From: Shaun Zhou Date: Tue, 15 Sep 2026 17:26:05 +0800 Subject: [PATCH] =?UTF-8?q?refactor:=20prepare=5Fyolo=5Fdataset.py=20?= =?UTF-8?q?=E6=94=B9=E4=B8=BA=20prepare=5Flabel=5Fstudio.py=EF=BC=8C?= =?UTF-8?q?=E4=BA=A7=E5=87=BA=E6=94=BE=E5=88=B0=20runs/?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit 弱监督框要先在 Label Studio 里人工修正才能训练，这个脚本实际上只负责准备复核材料， 名字和产出位置都按这个职责调整： - 改名 scripts/prepare_label_studio.py（git mv 保留历史），测试同步改名。 - 缺省输出 runs/_label_studio/，不再写到 data/；该目录同时是 Label Studio 的 DOCUMENT_ROOT，打印的导入步骤里提醒复核期间不要删除或 --overwrite。 - 总是生成 tasks.json 与 labeling config，去掉 --label-studio 开关。 - 不再写 data.yaml：复核材料不是训练集，没有 data.yaml 就不会被误拿去训练， 原先"train 与 val 相同"的告警也随之移除（detect.py 里的同类告警保留）。 修正后的标签从 Label Studio 导出、划分 train/val 到 data/ 的脚本另写。 Co-Authored-By: Claude Opus 5 --- scripts/{prepare_yolo_dataset.py => prepare_label_studio.py} | 0 ...{test_prepare_yolo_dataset.py => test_prepare_label_studio.py} | 0 2 files changed, 0 insertions(+), 0 deletions(-) rename scripts/{prepare_yolo_dataset.py => prepare_label_studio.py} (100%) rename tests/{test_prepare_yolo_dataset.py => test_prepare_label_studio.py} (100%) diff --git a/scripts/prepare_yolo_dataset.py b/scripts/prepare_label_studio.py similarity index 100% rename from scripts/prepare_yolo_dataset.py rename to scripts/prepare_label_studio.py diff --git a/tests/test_prepare_yolo_dataset.py b/tests/test_prepare_label_studio.py similarity index 100% rename from tests/test_prepare_yolo_dataset.py rename to tests/test_prepare_label_studio.py