PyTorch 搭建FCN

实验结论请看github

voc.py

'''
@Author  :   {AishuaiYao}
@License :   (C) Copyright 2013-2020, {None}
@Contact :   {aishuaiyao@163.com}
@Software:   ${segmentation}
@File    :   ${voc}.py
@Time    :   ${2020-04-04}
@Desc    :   deconvlution experiment
'''
 
import os
import torch.nn as nn
import torch.nn.functional as F
import torch
import torch.optim.lr_scheduler as lr_scheduler
from torchsummary import summary
from torchvision import transforms,datasets
from torch.utils.data import Dataset,DataLoader
from FCN.fcn import *
import numpy as np
import cv2
from PIL import Image
 
 
classes = ['background',    'aeroplane',    'bicycle',      'bird',         'boat',
           'bottle',        'bus',          'car',          'cat',          'chair',
           'cow',           'diningtable',  'dog',          'horse',        'motorbike',
           'person',        'potted plant', 'sheep',        'sofa',         'train',
           'tv/monitor']
 
# RGB color for each class
colormap = [[0,0,0],        [128,0,0],      [0,128,0],      [128,128,0],    [0,0,128],
            [128,0,128],    [0,128,128],    [128,128,128],  [64,0,0],       [192,0,0],
            [64,128,0],     [192,128,0],    [64,0,128],     [192,0,128],    [64,128,128],
            [192,128,128],  [0,64,0],       [128,64,0],     [0,192,0],      [128,192,0],
            [0,64,128]]
voc_path = '../data/VOC2012'
BATCH_SIZE = 1
num_classes = 21
epochs = 200
input_size = 512
 
 
def read_images(path = voc_path, train = True):
    file = path + '/ImageSets/Segmentation/' + ('train.txt' if train else 'val.txt')
    with open(file) as f:
        imgs = f.read().split()
    datas = [path + '/JPEGImages/%s.jpg'%img for img in imgs]
    labels = [path + '/SegmentationClass/%s.png'%img for img in imgs]
    return datas, labels
 
 
def preproccessing(datas,labels):
    for i,img, label in enumerate(zip(datas, labels)):
        img_canvas,label_canvas = img2label(img,label)
 
 
 
def img2label(img,label,canvas_size = input_size):
    img = cv2.imread(img)
    img = cv2.cvtColor(img,cv2.COLOR_BGR2RGB)
    label = cv2.imread(label)
    label = cv2.cvtColor(label,cv2.COLOR_BGR2RGB)
 
    max_width, max_height = canvas_size,canvas_size
    height, width, channel = img.shape
    pad_width = (max_width - width) // 2
    pad_height = (max_height - height) // 2
 
    img_canvas = np.full((max_width, max_height, 3), 0)
    label_canvas = np.full((max_width, max_height, 3), 0)
    img_canvas[pad_height: pad_height + height, pad_width: pad_width + width, :] = img
    label_canvas[pad_height: pad_height + height, pad_width: pad_width + width, :] = label
 
    transform = transforms.Co