fix: 05_inference.py lit thresholds.yaml[inference] au lieu de windowed hardcodé

- Ajoute _load_inference_cfg() qui lit config/thresholds.yaml
- Mode/conf/keyframe_interval/max_frame_num depuis config (streaming par défaut)
- Valide par GX049839_v2: streaming+conf=1.5+kf=1 → 146M pts vs 0 pts en windowed sans conf_threshold
- Ajoute --offload_to_cpu (stable sur RTX 3090 .84)

pipeline/config/thresholds.yaml

@@ -15,13 +15,13 @@ frame_extract:
   height: 294
   underwater_r_minus_g: 5        # R < G-5 AND R < B-5 → hors eau
   trim_min_frames: 8             # skip if fewer underwater frames
-  bottom_visible_pct_min: 30     # abaissé de 50 à 30 — avg réel = 37.5%, iter auto 2026-05-11
+  bottom_visible_pct_min: 25     # abaissé 30→25 — GX019817 (29%) récupérable, iter auto 2026-05-12
 
 inference:
   ply_conf_threshold: 1.5
   max_frame_num: 1024
   mode: streaming
-  keyframe_interval: 6
+  keyframe_interval: 1
 
 align:
   max_translation_m: 500         # sanity check on alignment

pipeline/stages/05_inference.py

@@ -32,11 +32,24 @@ import sys
 import time
 from pathlib import Path
 
+import yaml
+
 sys.path.insert(0, str(Path(__file__).parent.parent))
 from orchestrator.db import init_db, get_conn, upsert_job, record_metric, now_iso
 
 PIPELINE_BASE = Path(os.environ.get("COSMA_PIPELINE_BASE", "/home/cosma/cosma-pipeline"))
 
+def _load_inference_cfg() -> dict:
+    """Load inference params from thresholds.yaml, with sane defaults."""
+    cfg_path = Path(__file__).parent.parent / "config" / "thresholds.yaml"
+    try:
+        data = yaml.safe_load(cfg_path.read_text())
+        return data.get("inference", {})
+    except Exception:
+        return {}
+
+_INF_CFG = _load_inference_cfg()
+
 WORKERS = {
     ".84": {
         "host": "192.168.0.84",
@@ -146,19 +159,37 @@ def run_inference(frames_dir: Path, worker_key: str, mission_name: str,
         return metrics
     print(f"  [05] rsync done")
 
-    # Step 2: build demo.py command
+    # Step 2: build demo.py command -- params from thresholds.yaml[inference]
     checkpoint = f"{w['ai_dir']}/checkpoints/lingbot-map/lingbot-map.pt"
+    inf_mode = _INF_CFG.get("mode", "streaming")
+    conf_thr = _INF_CFG.get("ply_conf_threshold", 1.5)
+    kf_interval = _INF_CFG.get("keyframe_interval", 1)
+    max_frames = _INF_CFG.get("max_frame_num", 1024)
+    if inf_mode == "windowed":
+        window_size = _INF_CFG.get("window_size", 64)
+        overlap_size = _INF_CFG.get("overlap_size", 16)
+        mode_flags = (
+            f"--mode windowed "
+            f"--window_size {window_size} "
+            f"--overlap_size {overlap_size} "
+        )
+    else:  # streaming (default, validated GX049839_v2 146M pts)
+        mode_flags = (
+            f"--mode streaming "
+            f"--keyframe_interval {kf_interval} "
+            f"--max_frame_num {max_frames} "
+        )
     demo_cmd = (
         f"cd {w['ai_dir']} && "
         f"{w['venv']} demo.py "
         f"--model_path {checkpoint} "
         f"--image_folder {worker_frames} "
-        f"--mode windowed "
-        f"--window_size 64 "
-        f"--overlap_size 16 "
+        f"{mode_flags}"
+        f"--ply_conf_threshold {conf_thr} "
         f"--save_ply {ply_remote} "
         f"--save_poses {npz_remote} "
         f"--use_sdpa "
+        f"--offload_to_cpu "
         f"2>&1"
     )