PINTO_model_zoo 
1. Introduction
１年前に記事にしたMobileNetV2-SSDLiteのトレーニング環境構築記事を超簡易仕様にリメイクしました。 GPU対応版の最新のDockerが導入されている段階から作業に着手すると、１５分ほどでトレーニングを開始できると思います。 ミスり要素はほぼ無いと思います。 学習が終わったらトレーニング済みモデル(.ckpt/.pb)の Integer Quantization を実施して Github へコミットする予定です。 トレーニング用データセットの前処理など、全ての処理はシェルスクリプト内に記載してありますので、気になる方は constants.sh あるいは prepare_checkpoint_and_dataset.sh あるいは retrain_detection_model.sh の内容をご覧ください。
2. Environment
- Ubuntu 18.04 x86_64, RAM 16GB, Geforce GTX1070
- Tensorflow-GPU v1.15.2
- Docker 19.03.5
- Pascal VOC 2012/2007 Dataset
- MobileNetV2-SSDLite
3. Procedure
$ DETECT_DIR=${HOME}/edgetpu/detection && mkdir -p $DETECT_DIR
$ cd $DETECT_DIR && nano Dockerfile
### Ubuntu 18.04
FROM tensorflow/tensorflow:1.15.2-gpu-py3
RUN apt-get update && \
DEBIAN_FRONTEND=noninteractive apt-get install -y \
protobuf-compiler python-pil python-lxml python-tk \
autoconf automake libtool curl make g++ unzip wget git nano \
libgflags-dev libgoogle-glog-dev liblmdb-dev libleveldb-dev \
libhdf5-serial-dev libhdf5-dev python3-opencv python-opencv \
python3-dev python3-numpy python3-skimage gfortran libturbojpeg \
python-dev python-numpy python-skimage python3-pip python-pip \
libboost-all-dev libopenblas-dev libsnappy-dev software-properties-common \
protobuf-compiler python-pil python-lxml python-tk libfreetype6-dev pkg-config \
libpng-dev libhdf5-100 libhdf5-cpp-100
# Get the tensorflow models research directory, and move it into tensorflow source folder to match recommendation of installation
RUN git clone https://github.com/tensorflow/models.git && \
mkdir -p /tensorflow && \
mv models /tensorflow
# Install the Tensorflow Object Detection API from here
# https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/installation.md
# Install object detection api dependencies
RUN pip3 install Cython && \
pip3 install contextlib2 && \
pip3 install pillow && \
pip3 install lxml && \
pip3 install jupyter && \
pip3 install matplotlib && \
pip2 install Cython && \
pip2 install contextlib2 && \
pip2 install pillow && \
pip2 install lxml && \
pip2 install jupyter && \
pip2 install matplotlib
# Get protoc 3.0.0, rather than the old version already in the container
RUN curl -OL "https://github.com/google/protobuf/releases/download/v3.0.0/protoc-3.0.0-linux-x86_64.zip" && \
unzip protoc-3.0.0-linux-x86_64.zip -d proto3 && \
mv proto3/bin/* /usr/local/bin && \
mv proto3/include/* /usr/local/include && \
rm -rf proto3 protoc-3.0.0-linux-x86_64.zip
# Install pycocoapi
RUN git clone --depth 1 https://github.com/cocodataset/cocoapi.git && \
cd cocoapi/PythonAPI && \
make -j8 && \
cp -r pycocotools /tensorflow/models/research && \
cd ../../ && \
rm -rf cocoapi
# Run protoc on the object detection repo
RUN cd /tensorflow/models/research && \
protoc object_detection/protos/*.proto --python_out=.
# Set the PYTHONPATH to finish installing the API
ENV PYTHONPATH $PYTHONPATH:/tensorflow/models/research:/tensorflow/models/research/slim
# Install wget (to make life easier below) and editors (to allow people to edit the files inside the container)
RUN apt-get update && \
apt-get install -y wget nano
ARG work_dir=/tensorflow/models/research
# Get object detection transfer learning scripts.
ARG scripts_link="http://storage.googleapis.com/cloud-iot-edge-pretrained-models/docker/obj_det_scripts.tgz"
RUN cd ${work_dir} && \
wget -O obj_det_scripts.tgz ${scripts_link} && \
tar zxvf obj_det_scripts.tgz
WORKDIR ${work_dir}
Ctrl + O
Ctrl + X
$ docker build - < Dockerfile --tag training_container
$ docker run \
--runtime=nvidia \
--name training_container \
-it \
--privileged \
-p 6006:6006 \
training_container \
/bin/bash
### Download constants.sh
$ curl -sc /tmp/cookie "https://drive.google.com/uc?export=download&id=1e74pfaoHqp6WiWwTGamzN7wNRTDoR3q3" > /dev/null
$ CODE="$(awk '/_warning_/ {print $NF}' /tmp/cookie)"
$ curl -Lb /tmp/cookie "https://drive.google.com/uc?export=download&confirm=${CODE}&id=1e74pfaoHqp6WiWwTGamzN7wNRTDoR3q3" -o constants.sh
### Download prepare_checkpoint_and_dataset.sh
$ curl -sc /tmp/cookie "https://drive.google.com/uc?export=download&id=1Rk2VRGBwZnlcfl3DmryPMbNpICJ2R-vr" > /dev/null
$ CODE="$(awk '/_warning_/ {print $NF}' /tmp/cookie)"
$ curl -Lb /tmp/cookie "https://drive.google.com/uc?export=download&confirm=${CODE}&id=1Rk2VRGBwZnlcfl3DmryPMbNpICJ2R-vr" -o prepare_checkpoint_and_dataset.sh
### Download retrain_detection_model.sh
$ curl -sc /tmp/cookie "https://drive.google.com/uc?export=download&id=1MbjoqxBG56aDyvGxBbJd1QCbGo4nDKnm" > /dev/null
$ CODE="$(awk '/_warning_/ {print $NF}' /tmp/cookie)"
$ curl -Lb /tmp/cookie "https://drive.google.com/uc?export=download&confirm=${CODE}&id=1MbjoqxBG56aDyvGxBbJd1QCbGo4nDKnm" -o retrain_detection_model.sh
constants.sh
#!/bin/bash
declare -A ckpt_link_map
declare -A ckpt_name_map
declare -A config_filename_map
ckpt_link_map["mobilenet_v2_ssdlite"]="http://download.tensorflow.org/models/object_detection/ssdlite_mobilenet_v2_coco_2018_05_09.tar.gz"
ckpt_name_map["mobilenet_v2_ssdlite"]="ssdlite_mobilenet_v2_coco_2018_05_09"
config_filename_map["mobilenet_v2_ssdlite-true"]="pipeline_mobilenet_v2_ssdlite_retrain_whole_model.config"
INPUT_TENSORS='normalized_input_image_tensor'
OUTPUT_TENSORS='TFLite_Detection_PostProcess,TFLite_Detection_PostProcess:1,TFLite_Detection_PostProcess:2,TFLite_Detection_PostProcess:3'
OBJ_DET_DIR="$PWD"
#rm -rf "${OBJ_DET_DIR}/learn"
mkdir -p "${OBJ_DET_DIR}/learn"
LEARN_DIR="${OBJ_DET_DIR}/learn"
#rm -rf "${LEARN_DIR}/data"
#rm -rf "${LEARN_DIR}/ckpt"
#rm -rf "${LEARN_DIR}/train"
#rm -rf "${LEARN_DIR}/models"
mkdir -p "${LEARN_DIR}/data"
mkdir -p "${LEARN_DIR}/ckpt"
mkdir -p "${LEARN_DIR}/train"
mkdir -p "${LEARN_DIR}/models"
DATASET_DIR="${LEARN_DIR}/data"
CKPT_DIR="${LEARN_DIR}/ckpt"
TRAIN_DIR="${LEARN_DIR}/train"
OUTPUT_DIR="${LEARN_DIR}/models"
prepare_checkpoint_and_dataset.sh
#!/bin/bash
# Exit script on error.
set -e
# Echo each command, easier for debugging.
set -x
usage() {
cat << END_OF_USAGE
Downloads checkpoint and dataset needed for the tutorial.
--network_type Can be one of [mobilenet_v1_ssd, mobilenet_v2_ssd],
mobilenet_v1_ssd by default.
--train_whole_model Whether or not to train all layers of the model. false
by default, in which only the last few layers are trained.
--help Display this help.
END_OF_USAGE
}
network_type="mobilenet_v2_ssdlite"
train_whole_model="true"
while [[ $# -gt 0 ]]; do
case "$1" in
--network_type)
network_type=$2
shift 2 ;;
--train_whole_model)
train_whole_model=$2
shift 2;;
--help)
usage
exit 0 ;;
--*)
echo "Unknown flag $1"
usage
exit 1 ;;
esac
done
source "$PWD/constants.sh"
echo "PREPARING checkpoint..."
mkdir -p "${LEARN_DIR}"
ckpt_link="${ckpt_link_map[${network_type}]}"
ckpt_name="${ckpt_name_map[${network_type}]}"
cd "${LEARN_DIR}"
rm -rf "${CKPT_DIR}/${ckpt_name}"
wget -O "${CKPT_DIR}/${ckpt_name}.tar.gz" "$ckpt_link"
tar -zxvf "${CKPT_DIR}/${ckpt_name}.tar.gz" -C "${CKPT_DIR}"
rm "${CKPT_DIR}/${ckpt_name}.tar.gz"
rm -rf "${CKPT_DIR}/saved_model"
cp -r ${CKPT_DIR}/${ckpt_name}/* "${CKPT_DIR}/"
echo "CHOSING config file..."
cd "${OBJ_DET_DIR}"
curl -sc /tmp/cookie "https://drive.google.com/uc?export=download&id=1IoHkfxfkXSUM8nnevrb_dVct72fMJ7t2" > /dev/null
CODE="$(awk '/_warning_/ {print $NF}' /tmp/cookie)"
curl -Lb /tmp/cookie "https://drive.google.com/uc?export=download&confirm=${CODE}&id=1IoHkfxfkXSUM8nnevrb_dVct72fMJ7t2" -o pipeline.config
mv pipeline.config "${CKPT_DIR}"
echo "REPLACING variables in config file..."
sed -i "s%PATH_TO_BE_CONFIGURED/model.ckpt%${CKPT_DIR}/model.ckpt%g" "${CKPT_DIR}/pipeline.config"
sed -i "s%PATH_TO_BE_CONFIGURED/mscoco_label_map.pbtxt%${DATASET_DIR}/pascal_label_map.pbtxt%g" "${CKPT_DIR}/pipeline.config"
sed -i "s%PATH_TO_BE_CONFIGURED/mscoco_train.record%${DATASET_DIR}/pascal_train.record%g" "${CKPT_DIR}/pipeline.config"
sed -i "s%PATH_TO_BE_CONFIGURED/mscoco_val.record%${DATASET_DIR}/pascal_val.record%g" "${CKPT_DIR}/pipeline.config"
echo "PREPARING dataset"
rm -rf "${DATASET_DIR}"
mkdir -p "${DATASET_DIR}"
cd "${DATASET_DIR}"
# VOCtrainval_11-May-2012.tar <--- 1.86GB
curl -sc /tmp/cookie "https://drive.google.com/uc?export=download&id=1rATNHizJdVHnaJtt-hW9MOgjxoaajzdh" > /dev/null
CODE="$(awk '/_warning_/ {print $NF}' /tmp/cookie)"
curl -Lb /tmp/cookie "https://drive.google.com/uc?export=download&confirm=${CODE}&id=1rATNHizJdVHnaJtt-hW9MOgjxoaajzdh" -o VOCtrainval_11-May-2012.tar
# VOCtrainval_06-Nov-2007.tar <--- 460MB
curl -sc /tmp/cookie "https://drive.google.com/uc?export=download&id=1c8laJUn-aaWEhE5NlDwIdNv5ZdogUAcD" > /dev/null
CODE="$(awk '/_warning_/ {print $NF}' /tmp/cookie)"
curl -Lb /tmp/cookie "https://drive.google.com/uc?export=download&confirm=${CODE}&id=1c8laJUn-aaWEhE5NlDwIdNv5ZdogUAcD" -o VOCtrainval_06-Nov-2007.tar
# Extract the data.
tar -xvf VOCtrainval_11-May-2012.tar && rm VOCtrainval_11-May-2012.tar
tar -xvf VOCtrainval_06-Nov-2007.tar && rm VOCtrainval_06-Nov-2007.tar
echo "PREPARING label map..."
cd "${OBJ_DET_DIR}"
cp "object_detection/data/pascal_label_map.pbtxt" "${DATASET_DIR}"
echo "CONVERTING dataset to TF Record..."
protoc object_detection/protos/*.proto --python_out=.
python3 object_detection/dataset_tools/create_pascal_tf_record.py \
--label_map_path="${DATASET_DIR}/pascal_label_map.pbtxt" \
--data_dir=${DATASET_DIR}/VOCdevkit \
--year=merged \
--set=train \
--output_path="${DATASET_DIR}/pascal_train.record"
python3 object_detection/dataset_tools/create_pascal_tf_record.py \
--label_map_path="${DATASET_DIR}/pascal_label_map.pbtxt" \
--data_dir=${DATASET_DIR}/VOCdevkit \
--year=merged \
--set=val \
--output_path="${DATASET_DIR}/pascal_val.record"
retrain_detection_model.sh
#!/bin/bash
# Exit script on error.
set -e
# Echo each command, easier for debugging.
set -x
usage() {
cat << END_OF_USAGE
Starts retraining detection model.
--num_training_steps Number of training steps to run, 500 by default.
--num_eval_steps Number of evaluation steps to run, 100 by default.
--help Display this help.
END_OF_USAGE
}
num_training_steps=200000
while [[ $# -gt 0 ]]; do
case "$1" in
--num_training_steps)
num_training_steps=$2
shift 2 ;;
--num_eval_steps)
num_eval_steps=$2
shift 2 ;;
--help)
usage
exit 0 ;;
--*)
echo "Unknown flag $1"
usage
exit 1 ;;
esac
done
source "$PWD/constants.sh"
python3 object_detection/model_main.py \
--pipeline_config_path="${CKPT_DIR}/pipeline.config" \
--model_dir="${TRAIN_DIR}" \
--num_train_steps="${num_training_steps}" \
--num_eval_steps="${num_eval_steps}"
$ chmod +x constants.sh
$ chmod +x prepare_checkpoint_and_dataset.sh
$ chmod +x retrain_detection_model.sh
### Data preparation
$ ./prepare_checkpoint_and_dataset.sh
### Training
$ NUM_TRAINING_STEPS=1000000 && \
NUM_EVAL_STEPS=100 && \
rm -rf learn/train && \
./retrain_detection_model.sh \
--num_training_steps ${NUM_TRAINING_STEPS} \
--num_eval_steps ${NUM_EVAL_STEPS}
### Check learning status with Tensorboard
$ docker exec -it training_container /bin/bash
$ cd learn/train
$ tensorboard --logdir=.
＜Progress of learning＞
$ docker cp 07a048c6b1b8:/tensorflow/models/research/learn/train/model.ckpt-7044.data-00000-of-00001 .
$ docker cp 07a048c6b1b8:/tensorflow/models/research/learn/train/model.ckpt-7044.index .
$ docker cp 07a048c6b1b8:/tensorflow/models/research/learn/train/model.ckpt-7044.meta .
$ docker cp 07a048c6b1b8:/tensorflow/models/research/learn/train/graph.pbtxt .
$ docker cp 07a048c6b1b8:/tensorflow/models/research/learn/train/checkpoint .