YOLOv5模型介绍

Ultralytics YOLOv5 🚀 是由 Ultralytics 开发的尖端、最先进（SOTA）计算机视觉模型。YOLOv5 基于 PyTorch 框架构建，以其易用性、速度和准确性而闻名。它融合了广泛研究与开发中的见解和最佳实践，使其成为各种视觉 AI 任务的热门选择，包括目标检测、图像分割和图像分类。

（以上内容取自 Ultralytics YOLOv5 Github 仓库：https://github.com/ultralytics/yolov5）

模型下载与导出

1.Git下载YOLOv5；

sudo git clone https://github.com/ultralytics/yolov5

Cloning into 'yolov5'...

remote: Enumerating objects: 17968, done.

remote: Counting objects: 100% (115/115), done.

remote: Compressing objects: 100% (66/66), done.

2.安装依赖；

pip install onnx onnxruntime

Collecting onnx

Downloading onnx-1.17.0-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (16.0 MB)

|████████████████████████████████| 16.0 MB 201 kB/s

Collecting onnxruntime

3.导出模型；

python3 export.py --weights yolov5n.pt --include onnx --opset 11

export: data=data/coco128.yaml, weights=['yolov5n.pt'], imgsz=[640, 640], batch_size=1, device=cpu, half=False, inplace=False, keras=False, optimize=False, int8=False, per_tensor=False, dynamic=False, cache=, simplify=False, mlmodel=False, opset=11, verbose=False, workspace=4, nms=False, agnostic_nms=False, topk_per_class=100, topk_all=100, iou_thres=0.45, conf_thres=0.25, include=['onnx']

fatal: detected dubious ownership in repository at '/home/openexplorer/Desktop/yolov5'

To add an exception for this directory, call:

git config --global --add safe.directory /home/openexplorer/Desktop/yolov5

4.拷贝模型至“Desktop/dataset/”文件夹下；

模型验证与转换

1.启动 Open Explorer Docker环境

sudo docker run -it --rm -v /open_explorer -v ./dataset:/data/horizon_j5/data openexplorer/ai_toolchain_ubuntu_20_j5_cpu:v1.1.77-py38

2.模型验证

cd ..

hb_mapper checker --model-type onnx --model /data/horizon_j5/data/yolov5n.onnx --march bayes

/usr/local/lib/python3.8/dist-packages/paramiko/pkey.py:100: CryptographyDeprecationWarning: TripleDES has been moved to cryptography.hazmat.decrepit.ciphers.algorithms.TripleDES and will be removed from this module in 48.0.0.

"cipher": algorithms.TripleDES,

/usr/local/lib/python3.8/dist-packages/paramiko/transport.py:259: CryptographyDeprecationWarning: TripleDES has been moved to cryptography.hazmat.decrepit.ciphers.algorithms.TripleDES and will be removed from this module in 48.0.0.

3.图片转换

转换脚本如下：

import numpy as np
from PIL import Image

def generate_calib_data(img_path, output_path, target_size=(640, 640)):
  img = Image.open(img_path).convert('RGB')
  img_resized = img.resize(target_size, Image.BILINEAR)
  img_np = np.array(img_resized)
  img_np = img_np.transpose(2, 0, 1)
  img_np = np.expand_dims(img_np, axis=0)

运行测试如下：

python3 test.py

test.py:6: DeprecationWarning: BILINEAR is deprecated and will be removed in Pillow 10 (2023-07-01). Use Resampling.BILINEAR instead.

img_resized = img.resize(target_size, Image.BILINEAR)

Shape: (1, 3, 640, 640), Size: 1228800

4.模型转换

4.1编写yolov5n.yaml文件“Desktop/dataset/”文件夹下；

cat > /data/horizon_j5/data/yolov5n.yaml

model_parameters:
  onnx_model: '/data/horizon_j5/data/yolov5n.onnx'
  march: "bayes"
  layer_out_dump: False
  working_dir: '../model_output'
  output_model_file_prefix: 'yolov5n-det'
  remove_node_type: "Dequantize"
  node_info: {"/model.10/m/m.0/attn/Softmax": {'ON': 'BPU','InputType': 'int8','OutputType': 'int8'}}

4.2进行模型转换；

hb_mapper makertbin --config yolov5n.yaml --model-type onnx

/usr/local/lib/python3.8/dist-packages/paramiko/pkey.py:100: CryptographyDeprecationWarning: TripleDES has been moved to cryptography.hazmat.decrepit.ciphers.algorithms.TripleDES and will be removed from this module in 48.0.0.

"cipher": algorithms.TripleDES,

/usr/local/lib/python3.8/dist-packages/paramiko/transport.py:259: CryptographyDeprecationWarning: TripleDES has been moved to cryptography.hazmat.decrepit.ciphers.algorithms.TripleDES and will be removed from this module in 48.0.0.

"class": algorithms.TripleDES,

文章转载地平线论坛：【地平线征程5域控试用】（三）