tt-model authoring files (opt/rf-detr-v2 merged)

tt-model.yaml

@@ -94,7 +94,7 @@ card:
        {"label": "cat",    "label_id": 17, "score": 0.960, "box": [7.5, 54.4, 317.5, 470.6]},
        {"label": "remote", "label_id": 75, "score": 0.903, "box": [40.9, 72.7, 176.6, 117.7]}],
      "num_detections": 5, "image_size": {"width": 640, "height": 480}, "input_size": [560, 560],
-     "timing_ms": {"preprocess": 2.7, "inference": 43.4, "postprocess": 0.4, "total": 69.4}}
+     "timing_ms": {"decode": 5.8, "preprocess": 0.9, "inference": 11.7, "postprocess": 0.2, "total": 18.5}}
     ```
 
     - `box` is `[x1, y1, x2, y2]` in original image pixels; detections are sorted by score.
@@ -109,9 +109,9 @@ card:
 
     | Metric | Value |
     |---|---:|
-    | Detection-IoU agreement vs fp32 reference (demo image) | 98.67 (2 cat + 2 remote, per-object IoU 0.96–0.99) |
-    | Backbone feature-map PCC vs torch | 0.998–0.9998 |
-    | Inference, served over HTTP (warm, batch 1, 560×560) | ~41 ms device · ~50 ms end-to-end (~20 FPS) |
+    | Detection-IoU agreement vs fp32 reference (demo image) | 98.74 (2 cat + 2 remote, scores within 0.016 of the reference) |
+    | Backbone feature-map PCC vs torch | 0.999–0.9999 |
+    | Inference, served over HTTP (warm, batch 1, 560×560) | ~12 ms device · ~19 ms server-side incl. PNG decode (~85 FPS device-bound) |
 
     ### Caveats