yolov5得分socre远超过1

[luameows]

---

温馨提示：根据社区不完全统计，按照模板提问，可以加快回复和解决问题的速度

---

环境

• 【FastDeploy版本】： 说明具体的版本，如fastdeploy-linux-gpu-0.8.0
• 【编译命令】如果您是自行编译的FastDeploy，请说明您的编译方式（参数命令）
• 【系统平台】: Linux x64
• 【硬件】： Nvidia GPU 2080TI， CUDA 11.8 CUDNN 8.8
• 【编译语言】： C++

问题日志及出现问题的操作流程

• 附上详细的问题日志有助于快速定位分析
• 【模型精度问题】 跑yolov5-face-l模型，直接调用yolov5face的对象，输出score远大于1。 模型是yolov5官方模型yolov5l转onnx得到 测试用例 链接：https://pan.baidu.com/s/1e5dCIvRG0su3YkCYdmLitw?pwd=4ulv 提取码：4ulv

[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(362)::InitFromOnnx    number of streams:1.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(372)::InitFromOnnx    affinity:YES.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(384)::InitFromOnnx    Compile OpenVINO model on device_name:CPU.
[INFO] fastdeploy/runtime/runtime.cc(299)::CreateOpenVINOBackend        Runtime initialized with Backend::OPENVINO in Device::CPU.
DetectionResult: [xmin, ymin, xmax, ymax, score, label_id]
173.943359,204.884766, 429.113281, 553.580078, 240.251007, 2
173.945312,204.882812, 429.109375, 553.578125, 237.010391, 8
173.945312,204.882812, 429.113281, 553.578125, 208.403168, 4
173.945312,204.882812, 429.109375, 553.578125, 205.119736, 6
173.943573,204.884018, 429.113281, 553.579773, 200.753296, 0
173.945312,204.882812, 429.109375, 553.578125, 190.783569, 7
173.937500,204.890625, 429.109375, 553.578125, 189.610611, 9
173.945312,204.882812, 429.109375, 553.578125, 167.099899, 5
173.943359,204.883789, 429.113281, 553.580078, 141.013428, 1
173.945312,204.882812, 429.113281, 553.578125, 138.959610, 3
304.843750,957.109375, 325.203125, 984.781250, 2.805375, 9
304.843750,957.101562, 325.203125, 984.773438, 2.800912, 7
304.843750,957.101562, 325.203125, 984.773438, 2.789726, 5
304.843750,957.105469, 325.203125, 984.777344, 2.778498, 3
304.842773,957.105469, 325.202148, 984.775391, 2.773043, 1
304.841797,957.105469, 325.201172, 984.775391, 1.716787, 2
304.843750,957.101562, 325.203125, 984.773438, 1.714982, 8
304.843750,957.105469, 325.203125, 984.777344, 1.705539, 4
304.842651,957.104980, 325.201782, 984.775757, 1.695807, 0
304.843750,957.101562, 325.203125, 984.773438, 1.694840, 6
173.937500,204.890625, 429.109375, 553.578125, 0.808734, 10
298.859375,951.382812, 336.976562, 994.359375, 0.321541, 7
298.859375,951.375000, 336.984375, 994.359375, 0.321474, 9
298.859375,951.382812, 336.976562, 994.359375, 0.317842, 5
298.861328,951.382812, 336.977539, 994.358398, 0.315172, 1
298.863281,951.382812, 336.976562, 994.359375, 0.315045, 3

[luameows]

option选用trt作为后端，也是同样的问题。 模型我之前用tensorrt单独部署验证结果是正常