广州城市建设规划局网站,上海网站建设升级,Python能开发WordPress,在线建设网站 源代码目录
1. 前言
2.环境配置
(1) RK3588开发板Python环境
(2) PC转onnx和rknn的环境
3.PT模型转onnx
4. ONNX模型转RKNN
6.测试结果 1. 前言 yolov8就不介绍了#xff0c;详细的请见YOLOV8详细对比#xff0c;本文章注重实际的使用#xff0c;从拿到yolov8的pt检测模型详细的请见YOLOV8详细对比本文章注重实际的使用从拿到yolov8的pt检测模型怎样转onnx和rknn的使用配置好相应的环境就可以开心的做yolov8的检测啦
2.环境配置
(1) RK3588开发板Python环境 numpy 1.26.0 onnxruntime 1.16.0 opencv-python 4.8.0.76 rknn-toolkit-lite2 1.5.2
(2) PC转onnx和rknn的环境 Python 3.10 ultralytics 8.0.147 rknn_toolkit2 1.5.2 torch 2.0.1 torchvision0.15.2
相关的库缺少什么就安装很简单的
3.PT模型转onnx 根据自己模型路径设置model_path输入的shape也自己设置我这里设置的是640*640这样就得到了onnx模型按道理是可以做推理了
from ultralytics import YOLO# pip install ultralyticsinput_width 640
input_height 640
model_path ./models/yolov8x.ptmodel YOLO(model_path)
model.export(formatonnx, imgsz[input_height, input_width], opset12)
4. ONNX模型转RKNN rknn用FP16用INT8设置了但是量化效果精度有误差用混合量化或者增加量化数据集都可以试试(我偷懒就不尝试了)
from rknn.api import RKNNONNX_MODEL yolov8n-sim.onnx
RKNN_MODEL yolov8n-sim-fp16.rknn
IMG_PATH ./bus.jpg
DATASET ./dataset.txtif __name__ __main__:# Create RKNN objectrknn RKNN()if not os.path.exists(ONNX_MODEL):print(model not exist)exit(-1)# pre-process configprint(-- Config model)rknn.config(mean_values[[0, 0, 0]], std_values[[1, 1, 1]], target_platformrk3588)print(done)# Load ONNX modelprint(-- Loading model)ret rknn.load_onnx(modelONNX_MODEL)if ret ! 0:print(Load yolov8 failed!)exit(ret)print(done)# Build modelprint(-- Building model)ret rknn.build(do_quantizationFalse, datasetDATASET)if ret ! 0:print(Build yolov8 failed!)exit(ret)print(done)# Export RKNN modelprint(-- Export RKNN model)ret rknn.export_rknn(RKNN_MODEL)if ret ! 0:print(Export yolov8rknn failed!)exit(ret)print(done)# Ret rknn.accuracy_analysis(inputs[./bus.jpg])# if ret ! 0:# print(Accuracy analysis failed!)# exit(ret)# print(done)
5. RK3588推理 代码使用的是https://github.com/ibaiGorordo/ONNX-YOLOv8-Object-DetectionibaiGorord博主的基础代码在YOLOv8中集成了rknn模型的推理
import cv2
from yolov8 import YOLOv8# Initialize yolov8 object detector
# model_path ./models/yolov8n-sim.onnx
model_path ./models/yolov8n-sim-fp16.rknn
yolov8_detector YOLOv8(model_path, conf_thres0.3, iou_thres0.5)
# Read image
img cv2.imread(./bus.jpg)
# Detect Objects
boxes, scores, class_ids yolov8_detector(img)
# Draw detections
combined_img yolov8_detector.draw_detections(img)
cv2.imwrite(./detected_objects.png, combined_img)6.测试结果 onnx推理结果 rknn推理结果 使用的是yolov8n模型检测的速度加速了8倍量化成int8的话应该还要更加快 完整的代码在资源中下载
