mirror of https://github.com/alibaba/EasyCV.git
334 lines
8.3 KiB
Plaintext
334 lines
8.3 KiB
Plaintext
{
|
||
"cells": [
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "23af4ff9-2300-458b-9486-9278cb2abea5",
|
||
"metadata": {
|
||
"tags": []
|
||
},
|
||
"source": [
|
||
"# EasyCV度量学习\n",
|
||
"\n",
|
||
"本文将介绍如何使用EasyCV快速使用Resnet50进行度量学习模型的训练、推理"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "6c93bdca-6cb5-451e-92fd-653e33d2aa83",
|
||
"metadata": {
|
||
"tags": []
|
||
},
|
||
"source": [
|
||
"## 安装依赖包\n",
|
||
"\n",
|
||
"注: 在PAI-DSW docker中无需安装相关依赖，可跳过此部分 在本地notebook环境中执行\n"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "846076f6-ac34-49c9-aaff-e82aecd6336d",
|
||
"metadata": {},
|
||
"source": [
|
||
"1. 首先，安装pytorch和对应版本的torchvision，支持Pytorch1.5.1以上版本"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "259487df-b711-4457-ba0a-b32caf34079a",
|
||
"metadata": {
|
||
"tags": []
|
||
},
|
||
"outputs": [],
|
||
"source": [
|
||
"# install pytorch and torch vision\n",
|
||
"! conda install --yes pytorch==1.10.0 torchvision==0.11.0 -c pytorch"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "323f9718-927e-4377-aa80-cf5c398d54c5",
|
||
"metadata": {
|
||
"jp-MarkdownHeadingCollapsed": true,
|
||
"tags": []
|
||
},
|
||
"source": [
|
||
"2. 获取torch和cuda版本，安装对应版本的mmcv和nvidia-dali"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "57f87096-7801-4254-9dfd-cbdb4ab54cfc",
|
||
"metadata": {
|
||
"jp-MarkdownHeadingCollapsed": true,
|
||
"tags": []
|
||
},
|
||
"outputs": [],
|
||
"source": [
|
||
"import torch\n",
|
||
"import os\n",
|
||
"os.environ['CUDA']='cu' + torch.version.cuda.replace('.', '')\n",
|
||
"os.environ['Torch']='torch'+torch.version.__version__.replace('+PAI', '')\n",
|
||
"!echo $CUDA\n",
|
||
"! echo $Torch"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "99e24dec-b747-4a23-a981-9aae14e433cf",
|
||
"metadata": {
|
||
"tags": []
|
||
},
|
||
"outputs": [],
|
||
"source": [
|
||
"# install some python deps\n",
|
||
"! pip install mmcv-full==1.4.4 -f https://download.openmmlab.com/mmcv/dist/${CUDA}/${Torch}/index.html\n",
|
||
"! pip install http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/third_party/nvidia_dali_cuda100-0.25.0-1535750-py3-none-manylinux2014_x86_64.whl"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "6166dd4e-0e31-4061-9229-5f79968732e0",
|
||
"metadata": {
|
||
"jp-MarkdownHeadingCollapsed": true,
|
||
"tags": []
|
||
},
|
||
"source": [
|
||
"3. 安装EasyCV算法包"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "d55d2b68-b9d3-484a-abf6-561187153c54",
|
||
"metadata": {
|
||
"jp-MarkdownHeadingCollapsed": true,
|
||
"tags": []
|
||
},
|
||
"outputs": [],
|
||
"source": [
|
||
"pip install pai-easycv"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "d5e4d314-f59f-464c-950d-12c94b11020c",
|
||
"metadata": {
|
||
"tags": []
|
||
},
|
||
"source": [
|
||
"## 简单验证"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "80c51674-b1cc-4acf-a8e3-ed20306afebf",
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"from easycv.apis import *"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "0512498b-a900-4969-9563-67029a1009e2",
|
||
"metadata": {
|
||
"tags": []
|
||
},
|
||
"source": [
|
||
"## 运行环境要求\n",
|
||
"\n",
|
||
"PAI-Pytorch 1.7/1.8镜像， GPU机型 P100 or V100， 内存 32G"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "8df2d210-4bcf-4b41-a981-b264c41e91a7",
|
||
"metadata": {
|
||
"tags": []
|
||
},
|
||
"source": [
|
||
"## CUB200 度量学习"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "fce3f492-cfe0-4121-8f7d-877b31e3438b",
|
||
"metadata": {},
|
||
"source": [
|
||
"下面示例介绍如何利用[CUB200](http://www.vision.caltech.edu/datasets/cub_200_2011/)数据，使用ResNet50模型快速进行度量学习模型的训练评估、模型预测过程"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "6e769ad6-3ae0-4d5b-bf23-55673a239aa8",
|
||
"metadata": {},
|
||
"source": [
|
||
"### 数据准备\n",
|
||
"下载cub200数据，解压到`data/cub200`目录， 目录结构如下\n",
|
||
"\n",
|
||
"```text\n",
|
||
"data/cub200\n",
|
||
"├── images\n",
|
||
"├── images.txt\n",
|
||
"├── image_class_labels.txt\n",
|
||
"├── train_test_split.txt\n",
|
||
"```"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "62eeff3d-274d-4ff9-8123-8e11d97c0d38",
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"! mkdir -p data/ && wget https://s3.amazonaws.com/fast-ai-imageclas/CUB_200_2011.tgz && tar -xzf CUB_200_2011.tgz -C data/ && mv data/CUB_200_2011 data/cub200"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "ed28f2be-1519-4d81-9d03-42891d1e9bc7",
|
||
"metadata": {},
|
||
"source": [
|
||
"### 训练模型\n",
|
||
"下载训练配置文件"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "bd8b6d19-652c-4a58-89d3-38394dbb9a20",
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"! rm -rf cub_resnet50_jpg.py\n",
|
||
"!wget https://raw.githubusercontent.com/alibaba/EasyCV/master/configs/metric_learning/cub_resnet50_jpg.py"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "3d2ed375-f5ef-475e-9dc8-3386d328b8b0",
|
||
"metadata": {},
|
||
"source": [
|
||
"使用单卡gpu进行训练和验证集评估"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "9b6ec6da-d125-4ebe-b301-0d8cc8881512",
|
||
"metadata": {
|
||
"scrolled": true,
|
||
"tags": []
|
||
},
|
||
"outputs": [],
|
||
"source": [
|
||
"! CUDA_VISIBLE_DEVICES=0 python -m torch.distributed.launch --nproc_per_node=1 --master_port=29500 tools/train.py configs/metric_learning/cub_resnet50_jpg.py --work_dir work_dirs/metric_learning/cub/r50 --launcher pytorch --fp16"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "82b04e66-ef62-4d2e-963f-34015a588786",
|
||
"metadata": {},
|
||
"source": [
|
||
"### 导出模型\n",
|
||
"\n",
|
||
"模型训练完成，使用export命令导出模型进行推理，导出的模型包含推理时所需的预处理信息、后处理信息"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "1d4f7868-88aa-428c-9852-22e3eca45a07",
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"# 查看训练产生的pt文件\n",
|
||
"! ls work_dirs/metric_learning/cub/r50*"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "a5dc516f-fe35-4b1f-acbe-9ec96b72cc95",
|
||
"metadata": {},
|
||
"source": [
|
||
"ClsEvaluator_neck_top1_best.pth 是训练过程中产生的acc最高的pth,导出该模型"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "67aa8262-834a-4afc-a6cd-fd882a181371",
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"! python tools/export.py configs/metric_learning/cub_resnet50_jpg.py work_dirs/metric_learning/cub/r50/ClsEvaluator_neck_top1_best.pth work_dirs/metric_learning/cub/r50/best_export.pth"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "72d78ef4-0d90-4f3a-a1d9-c532e17d065b",
|
||
"metadata": {},
|
||
"source": [
|
||
"### 预测\n",
|
||
"下载测试图片"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "732fa5b4-fe0f-4a9b-ba19-d73d3ebd1736",
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"! wget http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/fine_grain_cls/cub_raw/images/001.Black_footed_Albatross/Black_Footed_Albatross_0001_796111.jpg"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"id": "cb3b24c2-7a19-4618-8bdf-42ebe1cdfeb3",
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"import cv2\n",
|
||
"from easycv.predictors.feature_extractor import TorchFeatureExtractor\n",
|
||
"\n",
|
||
"output_ckpt = 'work_dirs/metric_learning/cub/r50/best_export.pth'\n",
|
||
"tcls = TorchFeatureExtractor(output_ckpt)\n",
|
||
"\n",
|
||
"img = cv2.imread('Black_Footed_Albatross_0001_796111.jpg')\n",
|
||
"# input image should be RGB order\n",
|
||
"img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n",
|
||
"output = tcls.predict([img])\n",
|
||
"print(output)"
|
||
]
|
||
}
|
||
],
|
||
"metadata": {
|
||
"kernelspec": {
|
||
"display_name": "Python 3",
|
||
"language": "python",
|
||
"name": "python3"
|
||
},
|
||
"language_info": {
|
||
"codemirror_mode": {
|
||
"name": "ipython",
|
||
"version": 3
|
||
},
|
||
"file_extension": ".py",
|
||
"mimetype": "text/x-python",
|
||
"name": "python",
|
||
"nbconvert_exporter": "python",
|
||
"pygments_lexer": "ipython3",
|
||
"version": "3.6.13"
|
||
}
|
||
},
|
||
"nbformat": 4,
|
||
"nbformat_minor": 5
|
||
}
|