R6 YOLO11s FP16 ONNX: 77.01% mAP50 (was 69.22%)

conf_threshold 0.25→0.55: composite +30% (FP/img 6.26→1.09, mAP50 71.8→66.5%)

R5 FP16 ONNX: 69.22% mAP50, 5.4MB (beats R5 FP32 10.6MB on size, near-identical mAP)

v6-fp16: update chute_config.yml

v6-fp16: update miner.py

R5 COCO+VD-full+VD-testdev imgsz=1280: mAP@50=69.35%

R5 COCO+VD-full+VD-testdev imgsz=1280: mAP@50=69.35%

R5 epoch47 FP16 ONNX: 68.4% mAP50, 5.35MB, 10.2ms P50 (was R4 FP32: 63.4%, 10.7MB, 15.9ms)

Full VisDrone: mAP@50=58.76%

Full VisDrone: mAP@50=58.76%

Expanded VisDrone: mAP@50=51.21%

Expanded VisDrone: mAP@50=51.21%

VisDrone fine-tune: mAP@50=49.12%

VisDrone fine-tune: mAP@50=49.12%

Add main.py, pyproject.toml, update class_names.txt (80 COCO classes), update model_type.json and README

Upload folder using huggingface_hub

initial commit