introduction_cn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. 模型简介\n",
    "PaddleDetection中提出了全新的轻量级系列模型`PP-PicoDet`，在移动端具有卓越的性能，成为全新SOTA轻量级模型。\n",
    "\n",
    "PP-PicoDet模型有如下特点：\n",
    "\n",
    "- 🌟 更高的mAP: 第一个在1M参数量之内`mAP(0.5:0.95)`超越**30+**(输入416像素时)。\n",
    "- 🚀 更快的预测速度: 网络预测在ARM CPU下可达150FPS。\n",
    "- 😊 部署友好: 支持PaddleLite/MNN/NCNN/OpenVINO等预测库，支持转出ONNX，提供了C++/Python/Android的demo。\n",
    "- 😍 先进的算法: 我们在现有SOTA算法中进行了创新, 包括：ESNet, CSP-PAN, SimOTA等等。\n",
    "\n",
    "关于PP-Picodet的更多细节可以参考我们的[官方文档](https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/configs/picodet/README.md)。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. 模型效果\n",
    "PP-Picodet与其他轻量级模型的精度速度对比图如下所示：\n",
    "<div align=\"center\">\n",
    "  <img src=\"https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/docs/images/picodet_map.png\" width=500 />\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. 模型如何使用\n",
    "首先克隆PaddleDetection，并将数据集存放在`dataset/coco/`目录下面"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "%cd ~/work\n",
    "!git clone https://gitee.com/paddlepaddle/PaddleDetection\n",
    "%cd PaddleDetection"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.1 训练\n",
    "执行以下命令训练PP-Picodet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "vscode": {
     "languageId": "plaintext"
    }
   },
   "outputs": [],
   "source": [
    "# 单卡训练\n",
    "!CUDA_VISIBLE_DEVICES=0 python tools/train.py -c configs/picodet/picodet_s_320_coco_lcnet.yml --eval\n",
    "\n",
    "# 多卡训练\n",
    "!CUDA_VISIBLE_DEVICES=0,1,2,3 python -m paddle.distributed.launch --gpus 0,1,2,3 tools/train.py -c configs/picodet/picodet_s_320_coco_lcnet.yml --eval"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**注意:** \n",
    "- PicoDet所有模型均由4卡GPU训练得到，如果改变训练GPU卡数，需要按线性比例缩放学习率`base_lr`。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.2 部署\n",
    "PP-Picodet支持多种方式部署，具体可以参考[PP-Picodet部署](https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/configs/picodet/README.md#%E9%83%A8%E7%BD%B2)。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. 模型原理\n",
    "PP-Picodet的整体结构图如下所示：\n",
    "<div align=\"center\">\n",
    "  <img src=\"https://bj.bcebos.com/v1/paddledet/modelcenter/images/PP-Picodet-arch.png\" width=70% />\n",
    "</div>\n",
    "PP-Picodet由以下方法组成：\n",
    "- 增强的ShuffleNet-ESNet\n",
    "- CSP-PAN\n",
    "- SimOTA匹配策略\n",
    "\n",
    "更多细节可以参考我们的技术报告：https://arxiv.org/abs/2111.00902"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5. 注意事项\n",
    "**所有的命令默认运行在AI Studio的`jupyter`上, 如果运行在终端上，去掉命令开头的符号%或!**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 6. 相关论文及引用信息\n",
    "```\n",
    "@article{yu2021pp,\n",
    "  title={PP-PicoDet: A Better Real-Time Object Detector on Mobile Devices},\n",
    "  author={Yu, Guanghua and Chang, Qinyao and Lv, Wenyu and Xu, Chang and Cui, Cheng and Ji, Wei and Dang, Qingqing and Deng, Kaipeng and Wang, Guanzhong and Du, Yuning and others},\n",
    "  journal={arXiv preprint arXiv:2111.00902},\n",
    "  year={2021}\n",
    "}\n",
    "```"
   ]
  }
 ],
 "metadata": {
  "language_info": {
   "name": "python"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}