@echo off
REM ============================================================
REM  A2 (13/08/2026) - do nghiem thu lat A2 Analyzer da cu:
REM  chay lai TRON VOD rack_a1 (cung mau chuan da chot voi A1)
REM  de sinh overlay.mp4 + RMSE resim + co van tren 14 cu.
REM  Ban luu cua lan chay sinh so 13/08 (video_id bd1ac59b...).
REM
REM  Ly do chay (gate A2, BRIEF 13/08 sau nghiem thu A1):
REM    1. Co van "nghi khong phai cu danh": 3/3 cu ma (idx 7/8/9),
REM       0/11 cu that bi oan.
REM    2. overlay.mp4 du 14 cu, phat duoc NGOAI app (H.264);
REM       khop mat < 1 duong kinh bi tren 3 cu that.
REM    3. RMSE hien o bang + JSON per cu, hai noi khop, ghi ro bo
REM       params (shotnet mac dinh, fallback analytic dan nhan).
REM    5. CSV export UTF-8 BOM du 14 hang + co van + RMSE.
REM
REM  Can chay truoc: Redis (docker compose) + run_cv_worker.bat
REM  (worker boot A2 co them JIT pooltool 40-90s nguoi / ~2s numba
REM  cache am, TRUOC heartbeat) + run_app_queue.bat (port 8000).
REM
REM  Video vao TAI DUNG file rack_5245_5805.mp4 da cat o lan do A1
REM  (byte-for-byte cung input, khong re-encode lai) — neu mat file
REM  thi cat lai bang lenh ffmpeg trong run_analyzer_a1_vod.bat.
REM
REM  Do duoc 13/08 (14 cu, video_id bd1ac59b06b24bf5b0b6e1a39f920ce2):
REM    A2.1 DAT - van no dung {7,8,9}, 0/11 cu that bi oan.
REM    A2.2 DAT - overlay 14/14 file H.264 (tong 11.2MB), khop mat
REM         cu 1/5/10: vong track TREN bi cai, X va cham dung cho.
REM    A2.3 DAT - RMSE bang == JSON ca 14 cu, bo "shotnet";
REM         gia tri 412-1603mm (sim2real gap: V0 model lech thang
REM         + phi basin - HIEN TRUNG THUC, khong sua); cu 8 null
REM         (khong motion_start de neo t=0 - chinh la tin hieu van).
REM    A2.5 DAT - CSV BOM utf-8-sig, 15 dong (header+14), co van
REM         co/khong dung 3 cu ma, tieng Viet nguyen ven.
REM    Doi chieu so analyze A1<->A2: KHONG LECH o nao tren 14 cu
REM         (numpy 2.4.4->2.3.5 do cai pooltool vao cv-env: vo hai).
REM    Thoi gian: scan 180.1s (9600 frame, dup 789) + analyze loi
REM         43.6s/14 cu + resim/overlay ~7.5s/cu -> 14 cu terminal
REM         sau ~148s ke tu het scan; wall ~6 phut (RTX 3070).
REM ============================================================
set PYTHONUTF8=1

REM 1) Upload vao Analyzer da cu (app + cv_worker phai dang chay):
curl.exe -s -F "video=@D:\Khoa luan\poolcoach-rl\datasets\bb9_pilot\rack_a1\rack_5245_5805.mp4;type=video/mp4" -F "corners=[[527.5,869.7],[1380.1,864.0],[1232.3,278.5],[673.0,272.8]]" http://localhost:8000/api/analyzer/videos

REM 2) Theo doi: GET http://localhost:8000/api/analyzer/shots?video=(id buoc 1)
REM 3) Xuat CSV/JSON: GET /api/analyzer/videos/(id)/export.csv | export.json
REM    Ket qua ben: poolcoach-rl\data\analyzer\(id)\ (JSON + overlay_XX.mp4)
pause