xtkuang/cmvr_edge_ai
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
main
cmvr_edge_ai/models/detection/people-talking-yolov8x/v2
History
xtkuang 9c0b2a3a8a refactor edge ai server and add onnx models
2026-07-22 15:17:34 +08:00
..
manifest.json refactor edge ai server and add onnx models 2026-07-22 15:17:34 +08:00
model.onnx refactor edge ai server and add onnx models 2026-07-22 15:17:34 +08:00
README.md refactor edge ai server and add onnx models 2026-07-22 15:17:34 +08:00

README.md

People Talking YOLOv8x ONNX v2

这是 people-talking-yolov8x@1 的静态 CPU FP32 ONNX 部署版本。

字段 值
Model ID people-talking-yolov8x@2
Backend onnxruntime-yolov8
输入 float32[1,3,640,640]
输出 float32[1,6,8400]，NMS 外置
ONNX opset 17
参数量/FLOPs 68,125,494 / 257.4 GFLOPs
文件大小 272,787,930 bytes
SHA256 b4387307dbdf1083c0229a12e833ccdc4a3ad0063d5bdb7908765072470e840b
来源 v1/best.pt

部署只选择 talking on phone；含义不清的 class 0 label 仅为保持类别编号一致而保留。 manifest.json 由运行时在 session 创建前校验。

model: people-talking-yolov8x@2
detect_labels: [talking on phone]
model_options:
  weights: models/detection/people-talking-yolov8x/v2/model.onnx
  providers: [CPUExecutionProvider]
  intra_op_threads: 1
  inter_op_threads: 1
  imgsz: 640

ONNX 化移除了边缘运行时的 Torch/Ultralytics 依赖，但没有降低网络参数量或 FLOPs， FP32 文件也约为 v1 PT 的两倍。合成图 CPU smoke 可运行，但它不是现场精度或实时性 验收；资源不足时应训练 n/s 级模型或使用现场校准集开展 INT8，而不是宣称格式转换已 让 YOLOv8x 轻量化。许可与数据限制继承 v1 模型卡。