scorevision: push artifact

README.md

@@ -2,41 +2,75 @@
 tags:
 - element_type:detect
 - model:yolov11-nano
-- object:person
+- object:vehicle
 manako:
-  description: >
-    YOLOv11-nano fine-tuned for ground-level CCTV person detection on SN44.
-    Trained on CrowdHuman (15k, dense crowds) + BDD100K street pedestrians.
-    Conf threshold raised to 0.35 to minimise false positives.
-  source: meaculpitt/Detect-Person
-  prompt_hints: null
+  description: YOLO11n vehicle detector for CCTV surveillance
   input_payload:
   - name: frame
     type: image
-    description: RGB frame (ground-level CCTV)
+    description: RGB frame
   output_payload:
   - name: detections
     type: detections
-    description: Bounding boxes for detected persons
-  evaluation_score: 0.5563
-  last_benchmark:
-    type: coco_val2017
-    ran_at: '2026-03-25T02:58:57+00:00'
-    result_path: null
+    description: List of detections
 ---
 
-# Detect-Person — SN44
+# Score Vision SN44 — VehicleDetect Miner
 
-YOLOv11-nano fine-tuned for ground-level CCTV person detection.
+**Wallet:** LukeTao | **Hotkey:** default | **UID:** 128 | **Netuid:** 44
 
-| Metric | Value |
-|--------|-------|
-| mAP@50 (COCO val2017) | 55.63% |
-| Precision (conf=0.35) | 56.86% |
-| Recall | 50.67% |
-| Baseline to beat | 37.55% |
-| Model size | 5.6 MB |
-| Input size | 1280×1280 |
+## Model
 
-**Training data**: CrowdHuman (15k) + BDD100K (3.2k pedestrians)
-**Validation**: COCO val2017 persons (2,693 images)
+| Property | Value |
+|---|---|
+| Architecture | YOLO11-nano |
+| Input size | 640×640 |
+| Model file | `weights.onnx` |
+| Size | ~11 MB (well under 30 MB limit) |
+| Framework | ONNX Runtime (CUDA EP) |
+| mAP\@50 | **63.05%** (COCO val2017, vehicle classes) |
+
+## Classes
+
+| Output ID | Class | COCO Index |
+|---|---|---|
+| 0 | car | 2 |
+| 1 | bus | 5 |
+| 2 | truck | 7 |
+| 3 | motorcycle | 3 |
+
+## Performance
+
+Measured on RTX 4090, COCO val2017 images (640×640 letterbox):
+
+| Metric | Value | Target |
+|---|---|---|
+| Mean FPS (CUDA) | ~371 | ≥ 30 |
+| Mean FPS (CPU)  | ~34  | ≥ 30 |
+| P95 latency (CUDA) | 2.83 ms | < 50 ms |
+| Inference (GPU) | 2.70 ms | — |
+
+## Output Format
+
+```json
+[{
+  "x": 320.5,
+  "y": 240.1,
+  "width": 150.0,
+  "height": 90.0,
+  "confidence": 0.91,
+  "class_id": 0,
+  "class": "car"
+}]
+```
+
+## Files
+
+| File | Purpose |
+|---|---|
+| `weights.onnx` | ONNX model (YOLO11-nano, opset 12) |
+| `main.py` | Inference runner (reads class_names.txt automatically) |
+| `class_names.txt` | One class name per line |
+| `model_type.json` | Model metadata |
+| `pyproject.toml` | Python package dependencies |
+| `README.md` | This file |