ScoreVision miner - manak0/Detect-Person

Inference artifact submitted for the ScoreVision public open-source track, element manak0/Detect-Person (person detection).

Files

File	Purpose
miner.py	Entry point loaded by the platform. Self-contained (no imports from other files in this repository).
weights.onnx	End-to-end detection model. Non-maximum suppression is part of the graph.
config.json	Element configuration and runtime post-processing parameters.
chute_config.yml	Chutes deployment configuration (base image, node selector, runtime).
weights.onnx.report.json	Export contract check report (input/output shapes, opset, in-graph NMS, class names, size).
README.md	This file.

Model contract

Item	Value
Input	static [1, 3, imgsz, imgsz], float32, RGB scaled to [0, 1]
Output	single tensor [1, 300, 6]; each row is x1, y1, x2, y2, conf, cls_id
In-graph NMS	yes
Classes	cls_id 0 -> person
Per-frame payload	{"frame_id": ..., "boxes": [...], "keypoints": []}

The image size imgsz and the runtime thresholds (conf, nms_iou, max_det) are read from config.json.

Notes

• Only miner.py is imported by the platform; it performs no intra-repository imports.
• The total repository size stays within the track limit.