周翔宇 xzhou
0 Followers · 0 Following
  • Joined on 2026-08-27
Repositories
1
 Projects Packages Public Activity Starred Repositories
xzhou pushed to main at xzhou/ModelTrainEval 2026-10-14 17:08:29 +08:00
f9a4032783 feat: train fixed-size classifiers and export class probabilities
xzhou pushed to main at xzhou/ModelTrainEval 2026-10-14 15:32:04 +08:00
63bab36147 fix: suppress normal pseudo boxes in classifier exports
xzhou pushed to main at xzhou/ModelTrainEval 2026-10-14 13:30:06 +08:00
219cf23abc feat: add MobileNetV3-Small classifier configuration
xzhou pushed to main at xzhou/ModelTrainEval 2026-10-14 12:00:51 +08:00
03f667276b feat: export classifier with YOLO OpenVINO interface
xzhou pushed to main at xzhou/ModelTrainEval 2026-10-13 15:58:04 +08:00
d42a89ed13 fix(scripts): update default TILE_SIZE to 128 in retrain_from_label_studio.sh
xzhou pushed to main at xzhou/ModelTrainEval 2026-10-13 14:54:05 +08:00
8be1eefec2 fix(configs): update image size and dataset path in film.yaml to 128x128
xzhou pushed to main at xzhou/ModelTrainEval 2026-10-11 16:42:15 +08:00
bf72560326 docs: add AGENTS.md for vision model training and evaluation workflow
xzhou pushed to main at xzhou/ModelTrainEval 2026-10-09 16:04:07 +08:00
6a8941c18a docs: 修正 clusters/ 是软链的说法
xzhou pushed to main at xzhou/ModelTrainEval 2026-10-08 14:57:18 +08:00
29ff7a2c9b docs: 清掉失效的 train.py 引用与 ONNX 导出依赖组
xzhou pushed to main at xzhou/ModelTrainEval 2026-09-30 15:11:53 +08:00
c2bd5301b4 feat(deploy): 模型归档统一为 film_<边长>_<时间戳>
xzhou pushed to main at xzhou/ModelTrainEval 2026-09-29 15:07:51 +08:00
bbd5fa167e wip: C++ demo 支持批量清单
xzhou pushed to main at xzhou/ModelTrainEval 2026-09-29 14:56:23 +08:00
d692ffdf6e wip: C++ 参考实现（待编译验证）
xzhou pushed to main at xzhou/ModelTrainEval 2026-09-24 17:50:55 +08:00
ada67466d4 feat(deploy): 不依赖 ultralytics 的 OpenVINO 推理脚本
xzhou pushed to main at xzhou/ModelTrainEval 2026-09-24 15:33:43 +08:00
ddd77653b5 fix(detect): 推理尺寸与训练尺寸不一致时报警
xzhou pushed to main at xzhou/ModelTrainEval 2026-09-23 16:03:01 +08:00
6e746b7351 feat(scripts): 重训后同时维护 current.pt 与 best.pt
xzhou pushed to main at xzhou/ModelTrainEval 2026-09-23 15:22:59 +08:00
fad1fe9fbb feat(scripts): 重训流程加入切片步骤
xzhou pushed to main at xzhou/ModelTrainEval 2026-09-23 15:13:20 +08:00
23562448ba fix(detect): 切片统计的口径
xzhou pushed to main at xzhou/ModelTrainEval 2026-09-23 15:12:37 +08:00
474b7b50a0 feat(detect): 推理侧改为切片，与固定 160 输入一致
xzhou pushed to main at xzhou/ModelTrainEval 2026-09-23 12:28:00 +08:00
2a92275317 feat(detect): 固定 160x160 输入，只用裁剪与镜像填充，禁止缩放
xzhou pushed to main at xzhou/ModelTrainEval 2026-09-22 13:36:27 +08:00
1684fddd7f feat(scripts): 重训脚本可跳过重启 ML backend