远程服务器 Linux 用cityscape训练DeepLabv3模型（Pytorch版）并用图像测试

参考
https://blog.csdn.net/qq_45389690/article/details/111591713?utm_medium=distribute.pc_relevant_download.none-task-blog-baidujs-2.nonecase&depth_1-utm_source=distribute.pc_relevant_download.none-task-blog-baidujs-2.nonecase
https://blog.csdn.net/weixin_41919571/article/details/107906066
代码
https://github.com/jfzhang95/pytorch-deeplab-xception

出现问题
ImportError: No module named pycocotools.coco
解决
https://blog.csdn.net/u011961856/article/details/77676461

https://blog.csdn.net/joejeanjean/article/details/78839318?utm_medium=distribute.pc_relevant.none-task-blog-baidujs_title-6&spm=1001.2101.3001.4242

https://blog.csdn.net/haiyonghao/article/details/80472713?utm_medium=distribute.pc_relevant.none-task-blog-baidujs_title-11&spm=1001.2101.3001.4242
一定要把PythonAPI目录下除setup.py之外的所有文件拷贝到pytorch-deeplab-xception-master文件夹下

出现问题
ImportError: No module named ‘Queue’
解决
https://blog.csdn.net/DarrenXf/article/details/82962412

出现问题
from utils.loss import SegmentationLosses
ImportError: No module named loss
参考
https://blog.csdn.net/Diliduluw/article/details/103742766
解决
在utils文件下 新建一个空白的__init__.py

出现问题
AttributeError: ‘module’ object has no attribute ‘kaiming_normal_’
参考
https://blog.csdn.net/songchunxiao1991/article/details/83104893?utm_medium=distribute.pc_relevant_t0.none-task-blog-BlogCommendFromMachineLearnPai2-1.control&depth_1-utm_source=distribute.pc_relevant_t0.none-task-blog-BlogCommendFromMachineLearnPai2-1.control

https://blog.csdn.net/Aug0st/article/details/42707709
然后 我删除了所有pyc文件 （主要是不敢更新Python27\Lib\urllib2.pyc文件 怕影响别的算法）
删除指令 find /dir -name “*.pyc” | xargs rm -rf
但并没用

然后
我又配了一个python3.5 pytorch0.4.1的环境

出现问题
RuntimeError: CUDA error: out of memory
解决
train.py中改batch-size的default=2

出现问题
ValueError: Expected more than 1 value per channel when training, got input size [1, 256, 1, 1]

参考
https://blog.csdn.net/weixin_43925119/article/details/109755329
https://www.cnblogs.com/zmbreathing/p/pyTorch_BN_error.html
https://blog.csdn.net/sinat_39307513/article/details/87917537
https://blog.csdn.net/qq_42079689/article/details/102587401?utm_medium=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-2.control&depth_1-utm_source=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-2.control
https://blog.csdn.net/jining11/article/details/111478935?utm_medium=distribute.pc_relevant.none-task-blog-baidujs_title-2&spm=1001.2101.3001.4242
https://blog.csdn.net/qq_36321330/article/details/108954588?utm_medium=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-2.control&depth_1-utm_source=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-2.control
https://blog.csdn.net/qq_34124009/article/details/109100053?utm_medium=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-4.control&depth_1-utm_source=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-4.control
https://blog.csdn.net/qq_21230831/article/details/103711545?utm_medium=distribute.pc_relevant.none-task-blog-OPENSEARCH-7.control&depth_1-utm_source=distribute.pc_relevant.none-task-blog-OPENSEARCH-7.control
https://blog.csdn.net/lrs1353281004/article/details/108262018?utm_medium=distribute.pc_relevant.none-task-blog-baidujs_title-7&spm=1001.2101.3001.4242
模型中含有nn.BatchNorm层，训练时需要batch_size大于1，来计算当前batch的running mean and std，数据数量除以batch_size后刚好余1时就会报错。

解决
改batch_size 使其除完不余1，但我改完5之后内存又不行了，报错。
于是 我去找了/usr/local/lib/python3.5/dist-packages/torch/utils/data/dataloader.py文件，改里面drop_last=True。
因为远程服务器不能直接打开这个文件，所以粘过来又拷过去的。

跑的好慢啊！！！

跑完了，运行demo测试一下

import argparse
import os
import numpy as np
import time

from modeling.deeplab import *
from dataloaders import custom_transforms as tr
from PIL import Image
from torchvision import transforms
from dataloaders.utils import  *
from torchvision.utils import make_grid, save_image

def main():

    parser = argparse.ArgumentParser(description="PyTorch DeeplabV3Plus Training")
    parser.add_argument('--in-path', type=str, required=True, help='image to test')
    parser.add_argument('--out-path', type=str, required=True, help='mask image to save')
    parser.add_argument('--backbone', type=str, default='xception',
                        choices=['resnet', 'xception', 'drn', 'mobilenet'],
                        help='backbone name (default: xception)')
    parser.add_argument('--ckpt', type=str, default='deeplab-xception.pth',
                        help='saved model')
    parser.add_argument('--out-stride', type=int, default=16,
                        help='network output stride (default: 8)')
    parser.add_argument('--no-cuda', action='store_true', default=False,
                        help='disables CUDA training')
    parser.add_argument('--gpu-ids', type=str, default='0',
                        help='use which gpu to train, must be a \
                        comma-separated list of integers only (default=0)')
    parser.add_argument('--dataset', type=str, default='cityscapes',
                        choices=['pascal', 'coco', 'cityscapes','invoice'],
                        help='dataset name (default: cityscapes)')
    parser.add_argument('--crop-size', type=int, default=513,
                        help='crop image size')
    parser.add_argument('--num_classes', type=int, default=19,
                        help='crop image size')
    parser.add_argument('--sync-bn', type=bool, default=None,
                        help='whether to use sync bn (default: auto)')
    parser.add_argument('--freeze-bn', type=bool, default=False,
                        help='whether to freeze bn parameters (default: False)')

    args = parser.parse_args()
    args.cuda = not args.no_cuda and torch.cuda.is_available()
    if args.cuda:
        try:
            args.gpu_ids = [int(s) for s in args.gpu_ids.split(',')]
        except ValueError:
            raise ValueError('Argument --gpu_ids must be a comma-separated list of integers only')

    if args.sync_bn is None:
        if args.cuda and len(args.gpu_ids) > 1:
            args.sync_bn = True
        else:
            args.sync_bn = False
    model_s_time = time.time()
    model = DeepLab(num_classes=args.num_classes,
                    backbone=args.backbone,
                    output_stride=args.out_stride,
                    sync_bn=args.sync_bn,
                    freeze_bn=args.freeze_bn)

    ckpt = torch.load(args.ckpt, map_location='cpu')
    model.load_state_dict(ckpt['state_dict'])
    model = model.cuda()
    model_u_time = time.time()
    model_load_time = model_u_time-model_s_time
    print("model load time is {}".format(model_load_time))

    composed_transforms = transforms.Compose([
        tr.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)),
        tr.ToTensor()])
    for name in os.listdir(args.in_path):
        s_time = time.time()
        image = Image.open(args.in_path+"/"+name).convert('RGB')

        # image = Image.open(args.in_path).convert('RGB')
        target = Image.open(args.in_path+"/"+name).convert('L')
        sample = {
     'image': image, 'label': target}
        tensor_in = composed_transforms(sample)['image'].unsqueeze(0)

        model.eval()
        if args.cuda:
            tensor_in = tensor_in.cuda()
        with torch.no_grad():
            output = model(tensor_in)

        grid_image = make_grid(decode_seg_map_sequence(torch.max(output[:3], 1)[1].detach().cpu().numpy()),
                                3, normalize=False, range=(0, 255))
        save_image(grid_image,args.out_path+"/"+"{}_mask.png".format(name[0:-4]))
        u_time = time.time()
        img_time = u_time-s_time
        print("image:{} time: {} ".format(name,img_time))
        # save_image(grid_image, args.out_path)
        # print("type(grid) is: ", type(grid_image))
        # print("grid_image.shape is: ", grid_image.shape)
    print("image save in out_path.")
if __name__ == "__main__":
   main()

# python demo.py --in-path your_file --out-path your_dst_file --ckpt /home/ubuntu/pytorch-deeplab-xception-master/run/cityscapes/deeplab-xception/model_best.pth.tar --backbone xception

代码里面记得改成你自己用的模型，数据集之类的。

我的本意是用 这个用cityscape训练好的模型 去分割我自己的街景图，但我发现分割出来的图 颜色完全和cityscape不一样，

然后我发现 我的原图是jpg，于是我就将jpg转为png（因为cityscape的图是png）

import os
import sys

from PIL import Image

input_folder=r'C:/Users/wo/Desktop/yaogaishuju/human8'  #源文件夹，包含.png格式图片
output_folder=r'C:/Users/wo/Desktop/yaogaishuju/human8png'# 输出文件夹
#training_data=[]    
a=[]
for root, dirs, files in os.walk(input_folder):
    for filename in (x for x in files if x.endswith('.jpg')):
        filepath = os.path.join(root, filename) 
        
        object_class = filename.split('.')[0]
        a.append(object_class)
    print(a)
    
for i in a:
    old_path=input_folder+"\\"+str(i)+'.jpg'
    new_path=output_folder+"\\"+str(i)+'.png'
    img=Image.open(old_path)
    img.save(new_path)

然后将dome.py里面的
image = Image.open(args.in_path+"/"+name).convert(‘RGB’)
改成
image = Image.open(args.in_path+"/"+name).convert(‘RGBA’)

果然 就出错
ValueError: operands could not be broadcast together with shapes (576,768,4) (3,) (576,768,4)
维度不一样
要改的太多了，我又换回’RGB’了
合着白转成png了，输入的都是RGB三通道，
分割的图的颜色一点变化都没有。

有没有大神告诉我 为什么用分割网络得到的街景图 和cityscape的label颜色不一样？？？？？？


为什么我得到的分割颜色是这个鬼样子？？？？？

就你马离谱