# Beverage detector (cup / can / bottle) for manak0/Detect-beverage-detect - `weights.onnx`: single-scale (stride 8) YOLO-style detector, input 864x864 RGB, output [1, 7, N] (cx, cy, w, h in input pixels, then per-class sigmoid scores for cup, can, bottle). - `miner.py`: CPU onnxruntime with 2 intra-op threads, letterbox to the input size of the model, per-class NMS, then the set of boxes with the highest expected score is sent (the numbers are in the `esm` block of `config.json`). Nothing is sent when the best candidate scores below `gate` (an image without cups, cans or bottles). Deterministic numpy / onnxruntime only. - `chute_config.yml`: Chutes runtime with the package versions the miner was tested with. Latency of `predict_batch` on 2 pinned CPU cores (AMD EPYC 7542): 61 ms per 1024x1024 image on average (72 ms with PNG decoding, p95 75 ms).