pycharm 打开deeplabv3+ VOC数据集训练所需参数

曾经终败给现在 2024-04-17 06:25 99阅读 0赞

你需要的运行参数：https://github.com/tensorflow/models/blob/master/research/deeplab/local_test.sh
    也就是说复现的话你需要运行这个文件。在此之前你要安装好环境，并且将数据库整理好

#!/bin/bash
    # Copyright 2018 The TensorFlow Authors All Rights Reserved.
    #
    # Licensed under the Apache License, Version 2.0 (the "License");
    # you may not use this file except in compliance with the License.
    # You may obtain a copy of the License at
    #
    #     http://www.apache.org/licenses/LICENSE-2.0
    #
    # Unless required by applicable law or agreed to in writing, software
    # distributed under the License is distributed on an "AS IS" BASIS,
    # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    # See the License for the specific language governing permissions and
    # limitations under the License.
    # ==============================================================================
    #
    # This script is used to run local test on PASCAL VOC 2012. Users could also
    # modify from this script for their use case.
    #
    # Usage:
    #   # From the tensorflow/models/research/deeplab directory.
    #   sh ./local_test.sh
    #
    #
    
    # Exit immediately if a command exits with a non-zero status.
    set -e
    
    # Move one-level up to tensorflow/models/research directory.
    cd ..
    
    # Update PYTHONPATH.
    export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim
    export CUDA_VISIBLE_DEVICES=3,7
    # Set up the working environment.
    CURRENT_DIR=$(pwd)
    WORK_DIR="${CURRENT_DIR}/deeplab"
    
    # Run model_test first to make sure the PYTHONPATH is correctly set.
    #python "${WORK_DIR}"/model_test.py -v
    
    # Go to datasets folder and download PASCAL VOC 2012 segmentation dataset.
    DATASET_DIR="datasets"
    cd "${WORK_DIR}/${DATASET_DIR}"
    #sh download_and_convert_voc2012.sh
    
    # Go back to original directory.
    cd "${CURRENT_DIR}"
    
    # Set up the working directories.
    PASCAL_FOLDER="pascal_voc_seg"
    EXP_FOLDER="exp/train_on_trainval_set"
    INIT_FOLDER="${WORK_DIR}/${DATASET_DIR}/${PASCAL_FOLDER}/init_models"
    TRAIN_LOGDIR="${WORK_DIR}/${DATASET_DIR}/${PASCAL_FOLDER}/${EXP_FOLDER}/train"
    EVAL_LOGDIR="${WORK_DIR}/${DATASET_DIR}/${PASCAL_FOLDER}/${EXP_FOLDER}/eval"
    VIS_LOGDIR="${WORK_DIR}/${DATASET_DIR}/${PASCAL_FOLDER}/${EXP_FOLDER}/vis"
    EXPORT_DIR="${WORK_DIR}/${DATASET_DIR}/${PASCAL_FOLDER}/${EXP_FOLDER}/export"
    mkdir -p "${INIT_FOLDER}"
    mkdir -p "${TRAIN_LOGDIR}"
    mkdir -p "${EVAL_LOGDIR}"
    mkdir -p "${VIS_LOGDIR}"
    mkdir -p "${EXPORT_DIR}"
    
    # Copy locally the trained checkpoint as the initial checkpoint.
    TF_INIT_ROOT="http://download.tensorflow.org/models"
    TF_INIT_CKPT="deeplabv3_pascal_train_aug_2018_01_04.tar.gz"
    cd "${INIT_FOLDER}"
    #wget -nd -c "${TF_INIT_ROOT}/${TF_INIT_CKPT}"
    #tar -xf "${TF_INIT_CKPT}"
    cd "${CURRENT_DIR}"
    
    PASCAL_DATASET="${WORK_DIR}/${DATASET_DIR}/${PASCAL_FOLDER}/tfrecord"
    
    # Train 10 iterations.
    NUM_ITERATIONS=10
    echo ${INIT_FOLDER}
    echo "/deeplabv3_pascal_train_aug/model.ckpt"
    echo ${TRAIN_LOGDIR}
    echo ${PASCAL_DATASET}

参数

--logtostderr   
  --train\_split="trainval"   
  --model\_variant="xception\_65"   
  --atrous\_rates=6   
  --atrous\_rates=12   
  --atrous\_rates=18   
  --output\_stride=16   
  --decoder\_output\_stride=4   
  --train\_crop\_size="513,513"   
  --train\_batch\_size=4   
  --training\_number\_of\_steps="$\{NUM\_ITERATIONS\}"   
  --fine\_tune\_batch\_norm=true   
  --tf\_initial\_checkpoint="$\{INIT\_FOLDER\}/deeplabv3\_pascal\_train\_aug/model.ckpt"   
  --train\_logdir="$\{TRAIN\_LOGDIR\}"   
  --dataset\_dir="$\{PASCAL\_DATASET\}"

之后反应给我的绝对路径是：

/home/DATA/liutian/tmp/tfdeeplab/deeplab/datasets/pascal\_voc\_seg/init\_models/deeplabv3\_pascal\_train\_aug/model.ckpt  
/home/DATA/liutian/tmp/tfdeeplab/deeplab/datasets/pascal\_voc\_seg/exp/train\_on\_trainval\_set/train  
/home/DATA/liutian/tmp/tfdeeplab/deeplab/datasets/pascal\_voc\_seg/tfrecord  
这里tfdeeplab是我的项目目录。

数据库由于已经被build\_voc2012\_data.py转成tfrecord了

所以绝对路径参数：

之后反应给我的绝对路径是：

数据库由于已经被build\_voc2012\_data.py转成tfrecord了

最终直接将下面的参数复制到pycharm里面即可

--logtostderr --train\_split "train"  --model\_variant "xception\_65"   --atrous\_rates 6   --atrous\_rates 12   --atrous\_rates 18   --output\_stride 16   --decoder\_output\_stride 4   --train\_crop\_size "513,513"   --train\_batch\_size 4   --training\_number\_of\_steps 8  --fine\_tune\_batch\_norm true     --train\_logdir "/home/DATA/liutian/tmp/tfdeeplab/deeplab/datasets/pascal\_voc\_seg/exp/train\_on\_trainval\_set/train"   --dataset\_dir "/home/DATA/liutian/tmp/tfdeeplab/deeplab/datasets/pascal\_voc\_seg/tfrecord"

下面这个必须要删掉，否则会一直说找不到该文件

\--tf\_initial\_checkpoint "/home/DATA/liutian/tmp/tfdeeplab/deeplab/datasets/pascal\_voc\_seg/init\_models/deeplabv3\_pascal\_train\_aug/model.ckpt"

删掉之后就不会调用预训练的权重了

1.  The users should change atrous\_rates from \[6, 12, 18\] to \[12, 24, 36\] if setting output\_stride=8.
2.  The users could skip the flag, `decoder_output_stride`, if you do not want to use the decoder structure.

# Set to False if one does not want to re-use the trained classifier weights.
    flags.DEFINE_boolean('initialize_last_layer', True,
                         'Initialize the last layer.')

# When fine_tune_batch_norm=True, use at least batch size larger than 12
    # (batch size more than 16 is better). Otherwise, one could use smaller batch
    # size and set fine_tune_batch_norm=False.