|
|
|
#!/usr/bin/env python
|
|
|
|
|
|
|
|
# 目标检测模型PP-YOLO训练示例脚本
|
|
|
|
# 执行此脚本前,请确认已正确安装PaddleRS库
|
|
|
|
|
|
|
|
import os
|
|
|
|
|
|
|
|
import paddlers as pdrs
|
|
|
|
from paddlers import transforms as T
|
|
|
|
|
|
|
|
# 数据集存放目录
|
|
|
|
DATA_DIR = './data/sarship/'
|
|
|
|
# 训练集`file_list`文件路径
|
|
|
|
TRAIN_FILE_LIST_PATH = './data/sarship/train.txt'
|
|
|
|
# 验证集`file_list`文件路径
|
|
|
|
EVAL_FILE_LIST_PATH = './data/sarship/eval.txt'
|
|
|
|
# 数据集类别信息文件路径
|
|
|
|
LABEL_LIST_PATH = './data/sarship/labels.txt'
|
|
|
|
# 实验目录,保存输出的模型权重和结果
|
|
|
|
EXP_DIR = './output/ppyolo/'
|
|
|
|
|
|
|
|
# 下载和解压SAR影像舰船检测数据集
|
|
|
|
pdrs.utils.download_and_decompress(
|
|
|
|
'https://paddlers.bj.bcebos.com/datasets/sarship.zip', path='./data/')
|
|
|
|
|
|
|
|
# 定义训练和验证时使用的数据变换(数据增强、预处理等)
|
|
|
|
# 使用Compose组合多种变换方式。Compose中包含的变换将按顺序串行执行
|
|
|
|
# API说明:https://github.com/PaddlePaddle/PaddleRS/blob/develop/docs/apis/data.md
|
|
|
|
train_transforms = T.Compose([
|
|
|
|
# 读取影像
|
|
|
|
T.DecodeImg(),
|
|
|
|
# 随机裁剪,裁块大小在一定范围内变动
|
|
|
|
T.RandomCrop(),
|
|
|
|
# 随机水平翻转
|
|
|
|
T.RandomHorizontalFlip(),
|
|
|
|
# 对batch进行随机缩放,随机选择插值方式
|
|
|
|
T.BatchRandomResize(
|
|
|
|
target_sizes=[512, 544, 576, 608], interp='RANDOM'),
|
|
|
|
# 影像归一化
|
|
|
|
T.Normalize(
|
|
|
|
mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
|
|
|
|
T.ArrangeDetector('train')
|
|
|
|
])
|
|
|
|
|
|
|
|
eval_transforms = T.Compose([
|
|
|
|
T.DecodeImg(),
|
|
|
|
# 使用双三次插值将输入影像缩放到固定大小
|
|
|
|
T.Resize(
|
|
|
|
target_size=608, interp='CUBIC'),
|
|
|
|
# 验证阶段与训练阶段的归一化方式必须相同
|
|
|
|
T.Normalize(
|
|
|
|
mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
|
|
|
|
T.ArrangeDetector('eval')
|
|
|
|
])
|
|
|
|
|
|
|
|
# 分别构建训练和验证所用的数据集
|
|
|
|
train_dataset = pdrs.datasets.VOCDetDataset(
|
|
|
|
data_dir=DATA_DIR,
|
|
|
|
file_list=TRAIN_FILE_LIST_PATH,
|
|
|
|
label_list=LABEL_LIST_PATH,
|
|
|
|
transforms=train_transforms,
|
|
|
|
shuffle=True)
|
|
|
|
|
|
|
|
eval_dataset = pdrs.datasets.VOCDetDataset(
|
|
|
|
data_dir=DATA_DIR,
|
|
|
|
file_list=EVAL_FILE_LIST_PATH,
|
|
|
|
label_list=LABEL_LIST_PATH,
|
|
|
|
transforms=eval_transforms,
|
|
|
|
shuffle=False)
|
|
|
|
|
|
|
|
# 构建PP-YOLO模型
|
|
|
|
# 目前已支持的模型请参考:https://github.com/PaddlePaddle/PaddleRS/blob/develop/docs/intro/model_zoo.md
|
|
|
|
# 模型输入参数请参考:https://github.com/PaddlePaddle/PaddleRS/blob/develop/paddlers/tasks/object_detector.py
|
|
|
|
model = pdrs.tasks.det.PPYOLO(num_classes=len(train_dataset.labels))
|
|
|
|
|
|
|
|
# 执行模型训练
|
|
|
|
model.train(
|
|
|
|
num_epochs=10,
|
|
|
|
train_dataset=train_dataset,
|
|
|
|
train_batch_size=4,
|
|
|
|
eval_dataset=eval_dataset,
|
|
|
|
# 每多少个epoch存储一次检查点
|
|
|
|
save_interval_epochs=5,
|
|
|
|
# 每多少次迭代记录一次日志
|
|
|
|
log_interval_steps=4,
|
|
|
|
save_dir=EXP_DIR,
|
|
|
|
# 指定预训练权重
|
|
|
|
pretrain_weights='COCO',
|
|
|
|
# 初始学习率大小
|
|
|
|
learning_rate=0.0001,
|
|
|
|
# 学习率预热(learning rate warm-up)步数与初始值
|
|
|
|
warmup_steps=0,
|
|
|
|
warmup_start_lr=0.0,
|
|
|
|
# 是否启用VisualDL日志功能
|
|
|
|
use_vdl=True)
|