PaddleDetection系列-[1]-ppyoloe+训练自己的数据集

---

1.制作数据集

1.1. 用labelImg制作的数据集转换成PaddleDetection可以训练的coco数据集

labelImg制作的数据集包含JPEGImage图片目录和Annotationxml目录

1. 用脚本split_dataset.py用于划分训练集、测试集、验证集并制作成voc格式的

# -*- coding: UTF-8 -*-
import os
import random

base_dir = '/home/geekplusa/ai/datasets/'


def split_dataset():
    xmlfilepath = os.path.join(base_dir, 'Annotation')
    saveBasePath = os.path.join(base_dir, 'ImageSets/Main0')
    if not os.path.exists(saveBasePath):
        os.makedirs(saveBasePath)

    trainval_percent = 0.9
    train_percent = 0.9
    val_percent = 0.1

    temp_xml = os.listdir(xmlfilepath)
    total_xml = []
    for xml in temp_xml:
        if xml.endswith(".xml"):
            total_xml.append(xml)

    num = len(total_xml)
    list = range(num)
    tv = int(num * trainval_percent)
    tr = int(tv * train_percent)
    tval = int(tv * val_percent)

    trainval = random.sample(list, tv)
    train = random.sample(trainval, tr)
    val = random.sample(trainval, tval)

    print("train and val size", tv)
    print("traub suze", tr)
    ftrainval = open(os.path.join(saveBasePath, 'trainval.txt'), 'w')
    ftest = open(os.path.join(saveBasePath, 'test.txt'), 'w')
    ftrain = open(os.path.join(saveBasePath, 'train.txt'), 'w')
    fval = open(os.path.join(saveBasePath, 'val.txt'), 'w')

    for i in list:
        name = total_xml[i][:-4] + '\n'
        if i in trainval:
            ftrainval.write(name)
            if i in train:
                ftrain.write(name)
            else:
                fval.write(name)
        else:
            ftest.write(name)

    ftrainval.close()
    ftrain.close()
    fval.close()
    ftest.close()


if __name__ == '__main__':
    split_dataset()

2. 用PaddleDetection项目自带的tools/x2coco.py脚本生成coco格式的数据集,生成相对应的json文件.
   下面的例子是生成train.json的脚本例子

python tools/x2coco.py --dataset_type voc \
                       --image_input_dir '/home/geekplusa/ai/datasets/JPEGImage/ \
                       --voc_anno_dir '/home/geekplusa/ai/datasets/Annotation/ \
                       --voc_anno_list '/home/geekplusa/ai/datasets/ImageSets/Main0/train.txt \
                       --voc_label_list '/home/geekplusa/ai/datasets/label_list.txt \
                       --output_dir '/home/geekplusa/ai/datasets/ImageSets/coco \
                       --voc_out_name 'train.json' 

2.训练模型

2.1 修改相关的配置文件

1. coco_detection.yml修改成自己的数据

2.2 训练脚本

1. 单卡训练

python3 tools/train.py -c configs/ppyoloe/ppyoloe_plus_crn_x_80e_coco_my.yml --eval --amp

2. 多卡训练

nohup python3 -m paddle.distributed.launch --gpus 0,1,2,3 tools/train.py -c configs/ppyoloe/ppyoloe_plus_crn_x_80e_coco_my.yml --eval >logs/log_gkpw.log 2>&1 &

3.模型部署