yolov11 jetson nano 模型测试速度测试fp32 yolo11n.engine 67msfp32 change.engine 105msfp16 yolo11n.engine 47ms 提升20msfp16 change.engine 78ms 提升30msgpu占用ELSEjetson nano b01 环境配置yolov11改进的模型最好不要有什么可变性卷积速度测试fp32 yolo11n.engine 67ms0: 640x640110,67.1ms Speed:14.2ms preprocess,67.1ms inference,9.1ms postprocess per image at shape(1,3,640,640)0: 640x640110,67.5ms Speed:15.1ms preprocess,67.5ms inference,7.5ms postprocess per image at shape(1,3,640,640)0: 640x640110,65.7ms Speed:14.5ms preprocess,65.7ms inference,6.8ms postprocess per image at shape(1,3,640,640)0: 640x640110,68.1ms Speed:15.2ms preprocess,68.1ms inference,7.6ms postprocess per image at shape(1,3,640,640)0: 640x640110,68.3ms Speed:15.3ms preprocess,68.3ms inference,20.3ms postprocess per image at shape(1,3,640,640)0: 640x640110,69.0ms Speed:14.4ms preprocess,69.0ms inference,9.6ms postprocess per image at shape(1,3,640,640)0: 640x640110,66.5ms Speed:14.8ms preprocess,66.5ms inference,11.3ms postprocess per image at shape(1,3,640,640)0: 640x640110,66.4ms Speed:14.5ms preprocess,66.4ms inference,12.5ms postprocess per image at shape(1,3,640,640)0: 640x640110,65.7ms Speed:14.4ms preprocess,65.7ms inference,6.8ms postprocess per image at shape(1,3,640,640)fp32 change.engine 105ms0: 640x64013,135s,110,104.3ms Speed:18.1ms preprocess,104.3ms inference,9.9ms postprocess per image at shape(1,3,640,640)0: 640x640115s,16,110,103.8ms Speed:16.0ms preprocess,103.8ms inference,6.9ms postprocess per image at shape(1,3,640,640)0: 640x640115s,16,110,101.7ms Speed:16.7ms preprocess,101.7ms inference,9.8ms postprocess per image at shape(1,3,640,640)0: 640x64013,125s,110,105.7ms Speed:16.4ms preprocess,105.7ms inference,24.4ms postprocess per image at shape(1,3,640,640)0: 640x640105s,110,105.9ms Speed:14.6ms preprocess,105.9ms inference,11.4ms postprocess per image at shape(1,3,640,640)0: 640x640115s,110,104.3ms Speed:16.7ms preprocess,104.3ms inference,8.8ms postprocess per image at shape(1,3,640,640)0: 640x64013,125s,16,110,108.7ms Speed:15.3ms preprocess,108.7ms inference,6.4ms postprocess per image at shape(1,3,640,640)0: 640x640125s,110,101.8ms Speed:23.9ms preprocess,101.8ms inference,11.2ms postprocess per image at shape(1,3,640,640)0: 640x64095s,16,110,104.1ms Speed:14.9ms preprocess,104.1ms inference,8.5ms postprocess per image at shape(1,3,640,640)0: 640x640110,101.8ms Speed:14.7ms preprocess,101.8ms inference,6.9ms postprocess per image at shape(1,3,640,640)0: 640x640110,103.2ms Speed:14.8ms preprocess,103.2ms inference,7.6ms postprocess per image at shape(1,3,640,640)0: 640x640110,102.2ms Speed:16.1ms preprocess,102.2ms inference,16.2ms postprocess per image at shape(1,3,640,640)0: 640x640110,103.1ms Speed:14.4ms preprocess,103.1ms inference,11.4ms postprocess per image at shape(1,3,640,640)fp16 yolo11n.engine 47ms 提升20msSpeed:14.4ms preprocess,47.6ms inference,7.3ms postprocess per image at shape(1,3,640,640)0: 640x640110,47.9ms Speed:14.5ms preprocess,47.9ms inference,8.4ms postprocess per image at shape(1,3,640,640)0: 640x640110,48.6ms Speed:19.9ms preprocess,48.6ms inference,6.2ms postprocess per image at shape(1,3,640,640)0: 640x640110,47.5ms Speed:14.1ms preprocess,47.5ms inference,7.3ms postprocess per image at shape(1,3,640,640)0: 640x640110,47.8ms Speed:14.3ms preprocess,47.8ms inference,10.1ms postprocess per image at shape(1,3,640,640)0: 640x640110,50.0ms Speed:14.6ms preprocess,50.0ms inference,6.7ms postprocess per image at shape(1,3,640,640)0: 640x640110,49.1ms Speed:16.4ms preprocess,49.1ms inference,6.9ms postprocess per image at shape(1,3,640,640)0: 640x640110,47.7ms Speed:14.1ms preprocess,47.7ms inference,5.5ms postprocess per image at shape(1,3,640,640)0: 640x640110,47.9ms Speed:15.2ms preprocess,47.9ms inference,12.2ms postprocess per image at shape(1,3,640,640)fp16 change.engine 78ms 提升30ms0: 640x640110,78.9ms Speed:14.7ms preprocess,78.9ms inference,11.9ms postprocess per image at shape(1,3,640,640)0: 640x640110,78.7ms Speed:14.5ms preprocess,78.7ms inference,6.8ms postprocess per image at shape(1,3,640,640)0: 640x640110,79.8ms Speed:14.2ms preprocess,79.8ms inference,7.5ms postprocess per image at shape(1,3,640,640)0: 640x640110,81.6ms Speed:16.0ms preprocess,81.6ms inference,7.0ms postprocess per image at shape(1,3,640,640)0: 640x640110,79.1ms Speed:17.4ms preprocess,79.1ms inference,7.7ms postprocess per image at shape(1,3,640,640)0: 640x640110,78.9ms Speed:15.0ms preprocess,78.9ms inference,7.3ms postprocess per image at shape(1,3,640,640)0: 640x640(no detections),79.0ms Speed:14.5ms preprocess,79.0ms inference,4.2ms postprocess per image at shape(1,3,640,640)0: 640x640110,81.9ms Speed:16.0ms preprocess,81.9ms inference,8.5ms postprocess per image at shape(1,3,640,640)gpu占用fp16仅仅比fp32少占用10M gpu显存效果不明显ELSEjetson nano b01 环境配置yolov11改进的模型最好不要有什么可变性卷积tensorrt不支持所以也就无法转成.engine格式最好在jetson nano环境做onnx到engine的转换需要tensorrt的版本一致才可以
