deeplab script---python

Hi_Panda_CRL

于 2016-06-01 07:53:21 发布

阅读量1.5k

点赞数

分类专栏： Caffe

本文链接：https://blog.csdn.net/grief_of_the_nazgul/article/details/51553187

版权

该博客展示了如何使用Python脚本在deeplab框架下进行模型训练和测试。通过导入必要的库，设置实验参数，创建环境变量，并定义运行不同阶段的函数，实现对deeplab模型的训练、验证和保存操作。脚本中还包含了针对DenseCRF的选项以及参数搜索功能。

摘要由CSDN通过智能技术生成

import os,sys, subprocess
#get current folder and enter my_script folder
sys.path.insert(0, os.getcwd()+'/python/my_script/')

from tester import tester
from trainer import trainer
from crf_runner import crf_runner, grid_search
import tools

# MODIFY PATH for YOUR SETTING
EXP='voc12' #dataset
NET_ID='vgg128_noup' #model name
NUM_LABELS=21
YEAR = 'VOC2012'
DATA_ROOT='/media/ali/VOCdevkit/VOC2012'
#DATA_ROOT=subprocess.Popen('cd .. && pwd', stdout=subprocess.PIPE, shell=True).communicate()[0][:-1] + '/VOCdevkit/' + YEAR
OLD_ROOT=''#only change if you are changing the path to images
DEV_ID=0 #gpu id
LOAD_MAT_FILE = 1
       
train_set_SUFFIX='_aug'

train_set_STRONG='train'
#train_set_STRONG='train200'
#train_set_STRONG='train500'
#train_set_STRONG='train1000'
#train_se