tensorflow

tensorflow-1.0.0.sh

# Expected variables:
#
# STEPS
# MODEL_NAME
# BATCH_SIZE
# NUM_STEPS
# NUM_CLASSES
# UUID

echo "Detected arguments:"
for ARGUMENT in "$@"
do
    KEY=$(echo $ARGUMENT | cut -f1 -d=)
    VALUE=$(echo $ARGUMENT | cut -f2 -d=)
    export "$KEY"=$VALUE
    echo "$KEY"=$VALUE
done

EXEC_HOME=$(pwd)

if [ -z "${UUID}" ]; then
  # UUID=$(uuidgen)
  UUID=$(cat /proc/sys/kernel/random/uuid)
fi

echo ""
echo -n "Did you download the model? (y/n)  "
read ans1
echo -n "Does the pipeline.config has the batch_size inmediatly after or train_config? (y/n)  "
read ans2
echo -n "Does the pipeline.config has the tfrecord in the standard line number? (y/n)  "
read ans3

echo ""
echo "################### "
echo "UUID: "$UUID
echo "################### "

TF2_WORKSPACE=$(pwd)/workspace/$MODEL_NAME/$UUID

if [ -d "$TF2_WORKSPACE" ]; then
  echo "workspace already exist"
fi
mkdir -p $TF2_WORKSPACE

# download model if not exist

echo ""
echo "################### "
echo "model existence validation"
echo "################### "

if [ -d "$EXEC_HOME/models/$MODEL_NAME" ]
then
    echo "Model aready exist: $EXEC_HOME/models/$MODEL_NAME"
else
    echo "Model don't exist: $EXEC_HOME/models/$MODEL_NAME"
    # download urls
    curl --silent --output /dev/null https://raw.githubusercontent.com/tensorflow/models/master/research/object_detection/g3doc/tf2_detection_zoo.md -o $EXEC_HOME/models/tf2_detection_zoo.md
    # cat $EXEC_HOME/models/tf2_detection_zoo.md | grep -Eo "(http|https)://[a-zA-Z0-9./?=_%:-]*" | grep "download.tensorflow" > $EXEC_HOME/models/tf2_detection_zoo_urls.txt
    # cat  $EXEC_HOME/models/tf2_detection_zoo_urls.txt
    model_name_ocurrence_line=$(grep -n -m 1 "$MODEL_NAME" $EXEC_HOME/models/tf2_detection_zoo_urls.txt  | cut -f1 -d:)
    if [ -z "${model_name_ocurrence_line}" ]; then
      echo "Model $MODEL_NAME name was not found on official tensorflow2 models: https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2_detection_zoo.md"
      echo "Available models are:"
      cat $EXEC_HOME/models/tf2_detection_zoo_urls.txt
      echo "Choose one (without tar.gz) and use it as MODEL_NAME parameter and try again"
      exit 1
    fi
    echo ""
    model_download_url=$(sed $model_name_ocurrence_line"q;d" $EXEC_HOME/models/tf2_detection_zoo_urls.txt)
    echo "Downloading: "$model_download_url
    echo "Into: "$EXEC_HOME/models/$MODEL_NAME.tar.gz
    curl -silent --output /dev/null $model_download_url -o "$EXEC_HOME/models/$MODEL_NAME.tar.gz"
    # curl --silent --output /dev/null $model_download_url -o "$EXEC_HOME/models/$MODEL_NAME.tar.gz"
    echo "Decompressing"
    mkdir -p $EXEC_HOME/models/$MODEL_NAME
    tar -xzf "$EXEC_HOME/models/$MODEL_NAME.tar.gz" -C $EXEC_HOME/models/$MODEL_NAME --strip-components=1
    if [ "$?" -ne "0" ]; then
        echo "Error: downloading or decompressing model"
        exit 1
    fi
fi


if [[ $STEPS == *"0"* ]]; then
  echo ""
  echo "################### "
  echo "pipeline.config customization"
  echo "################### "

  cp $EXEC_HOME/models/$MODEL_NAME/pipeline.config $TF2_WORKSPACE

  # replace num_classes: 90 > num_classes: 1
  if [[ -n "$NUM_CLASSES" ]]
  then
    sed -i -e "s/num_classes: 90/num_classes: $NUM_CLASSES/g" $TF2_WORKSPACE/pipeline.config
    echo "success num_classes customization"
  fi

  # replace batch_size: 24 > batch_size: 8
  if [[ -n "$BATCH_SIZE" ]]
  then
    # sed -i -e 's/batch_size: 24/batch_size: 1/g' $TF2_WORKSPACE/pipeline.config
    batch_size_train_ocurrence_line=$(grep -n -m 1 'train_config' $TF2_WORKSPACE/pipeline.config  | cut -f1 -d:)
    batch_size_train_ocurrence_line=$((batch_size_train_ocurrence_line+1))
    sed -i $batch_size_train_ocurrence_line"s/.*/  batch_size: $BATCH_SIZE/" $TF2_WORKSPACE/pipeline.config
    echo "success batch_size customization"
  fi

  # num_steps
  if [[ -n "$NUM_STEPS" ]]
  then
    num_steps_ocurrence_line=$(grep -n -m 1 'num_steps' $TF2_WORKSPACE/pipeline.config  | cut -f1 -d:)
    sed -i $num_steps_ocurrence_line"s/.*/  num_steps: $NUM_STEPS/" $TF2_WORKSPACE/pipeline.config
    echo "success num_steps customization"
  fi

  # fine_tune_checkpoint:
  fine_tune_checkpoint_ocurrence_line=$(grep -n -m 1 'fine_tune_checkpoint:' $TF2_WORKSPACE/pipeline.config  | cut -f1 -d:)
  fine_tune_checkpoint_location=$EXEC_HOME/models/$MODEL_NAME/checkpoint/ckpt-0
  fine_tune_checkpoint_location=$(echo $fine_tune_checkpoint_location | sed 's_/_\\/_g')
  sed -i $fine_tune_checkpoint_ocurrence_line"s/.*/  fine_tune_checkpoint: \"$fine_tune_checkpoint_location\"/" $TF2_WORKSPACE/pipeline.config
  echo "success fine_tune_checkpoint customization"

  # label_map_path
  label_map_path_ocurrences_lines=$(grep -n 'label_map_path' $TF2_WORKSPACE/pipeline.config  | cut -f1 -d:)
  label_map_path_location=$EXEC_HOME/label_map.pbtxt
  label_map_path_location=$(echo $label_map_path_location | sed 's_/_\\/_g')
  for label_map_path_ocurrence_line in `echo $label_map_path_ocurrences_lines`; do
    sed -i $label_map_path_ocurrence_line"s/.*/  label_map_path: \"$label_map_path_location\"/" $TF2_WORKSPACE/pipeline.config
  done
  echo "success label_map_path customization"

  # train.record
  train_record_ocurrence_line=$(grep -n -m 1 'train_input_reader' $TF2_WORKSPACE/pipeline.config  | cut -f1 -d:)
  train_record_ocurrence_line=$((train_record_ocurrence_line+3))
  train_record_path_location=$TF2_WORKSPACE/data/train.record
  train_record_path_location=$(echo $train_record_path_location | sed 's_/_\\/_g')
  sed -i $train_record_ocurrence_line"s/.*/    input_path: \"$train_record_path_location\"/" $TF2_WORKSPACE/pipeline.config
  echo "success train.record customization"

  # test.record
  test_record_ocurrence_line=$(grep -n -m 1 'eval_input_reader' $TF2_WORKSPACE/pipeline.config  | cut -f1 -d:)
  test_record_ocurrence_line=$((test_record_ocurrence_line+3))
  test_record_path_location=$TF2_WORKSPACE/data/test.record
  test_record_path_location=$(echo $test_record_path_location | sed 's_/_\\/_g')
  sed -i $test_record_ocurrence_line"s/.*/    input_path: \"$test_record_path_location\"/" $TF2_WORKSPACE/pipeline.config
  echo "success test.record customization"

  # fine_tune_checkpoint_type
  fine_tune_checkpoint_type_ocurrence_line=$(grep -n -m 1 'fine_tune_checkpoint_type' $TF2_WORKSPACE/pipeline.config  | cut -f1 -d:)
  sed -i $fine_tune_checkpoint_type_ocurrence_line"s/.*/  fine_tune_checkpoint_type: \"detection\"/" $TF2_WORKSPACE/pipeline.config
  echo "success fine_tune_checkpoint_type customization"

fi
if [ "$?" -ne "0" ]; then
    echo "Error: pipeline.config customization"
    exit 1
fi


if [[ $STEPS == *"1"* ]]; then
  start_time=$(date +%s)
  echo ""
  echo "################### "
  echo "generate tfrecord"
  echo "################### "

  mkdir -p $TF2_WORKSPACE/data

  python3 generate_tfrecord.py -x $EXEC_HOME/images/train -l $EXEC_HOME/label_map.pbtxt -o $TF2_WORKSPACE/data/train.record
  python3 generate_tfrecord.py -x $EXEC_HOME/images/test -l $EXEC_HOME/label_map.pbtxt -o $TF2_WORKSPACE/data/test.record
fi
if [ "$?" -ne "0" ]; then
    bash elapsed_time.sh $start_time
    echo "Error: generate tfrecord"
    exit 1
else
  if [[ $STEPS == *"1"* ]]; then
    bash elapsed_time.sh $start_time
  fi
fi


if [[ $STEPS == *"2"* ]]; then
  start_time=$(date +%s)
  echo ""
  echo "################### "
  echo "train model"
  echo "################### "
  rm -rf $TF2_WORKSPACE/model/trained
  mkdir -p $TF2_WORKSPACE/model/trained
  python3 model_main_tf2.py --model_dir=$TF2_WORKSPACE/model/trained --pipeline_config_path=$TF2_WORKSPACE/pipeline.config
fi
if [ "$?" -ne "0" ]; then
  echo "Error: train model"
  bash elapsed_time.sh $start_time
  exit 1
else
  if [[ $STEPS == *"2"* ]]; then
    bash elapsed_time.sh $start_time
  fi
fi


if [[ $STEPS == *"3"* ]]; then
  start_time=$(date +%s)
  echo ""
  echo "################### "
  echo "export trained model"
  echo "################### "

  rm -rf $TF2_WORKSPACE/model/exported
  mkdir -p $TF2_WORKSPACE/model/exported
  python3 $EXEC_HOME/exporter_main_v2.py --input_type image_tensor --pipeline_config_path $EXEC_HOME/pipeline.config --trained_checkpoint_dir $TF2_WORKSPACE/model/trained --output_directory $TF2_WORKSPACE/model/exported
fi
if [ "$?" -ne "0" ]; then
  bash elapsed_time.sh $start_time
  echo "Error: export trained model"
  exit 1
else
  if [[ $STEPS == *"3"* ]]; then
    bash elapsed_time.sh $start_time
  fi
fi


if [[ $STEPS == *"4"* ]]; then
  start_time=$(date +%s)
  echo ""
  echo "################### "
  echo "object detection"
  echo "################### "
  python3 object_detection_exec_v2.py $TF2_WORKSPACE/model/exported $EXEC_HOME/label_map.pbtxt $EXEC_HOME/images/detection
fi
if [ "$?" -ne "0" ]; then
    bash elapsed_time.sh $start_time
    echo "Error: object detection"
    exit 1
else
  if [[ $STEPS == *"4"* ]]; then
    bash elapsed_time.sh $start_time
  fi
fi

tf1 - object_detection_tutorial.py

# coding: utf-8

# # Object Detection Demo
# Welcome to the object detection inference walkthrough!  This notebook will walk you step by step through the process of using a pre-trained model to detect objects in an image. Make sure to follow the [installation instructions](https://github.com/tensorflow/models/blob/master/object_detection/g3doc/installation.md) before you start.

# # Imports

# In[1]:


import numpy as np
import os
import six.moves.urllib as urllib
import sys
import tarfile
import tensorflow as tf
import zipfile

from collections import defaultdict
from io import StringIO
from matplotlib import pyplot as plt
from PIL import Image


# ## Env setup

# In[2]:


# This is needed to display the images.
# get_ipython().magic(u'matplotlib inline')
import matplotlib

# This is needed since the notebook is stored in the object_detection folder.
sys.path.append("..")


# ## Object detection imports
# Here are the imports from the object detection module.

# In[3]:


from object_detection.utils import label_map_util

from object_detection.utils import visualization_utils as vis_util


# # Model preparation

# ## Variables
#
# Any model exported using the `export_inference_graph.py` tool can be loaded here simply by changing `PATH_TO_CKPT` to point to a new .pb file.
#
# By default we use an "SSD with Mobilenet" model here. See the [detection model zoo](https://github.com/tensorflow/models/blob/master/object_detection/g3doc/detection_model_zoo.md) for a list of other models that can be run out-of-the-box with varying speeds and accuracies.

# In[4]:


# What model to download.
MODELS_HOME_DIR = sys.argv[1]
MODEL_NAME = sys.argv[2]
PATH_TO_LABELS = sys.argv[3]
PATH_TO_TEST_IMAGES_DIR = sys.argv[4]

MODEL_FILE = MODEL_NAME + '.tar.gz'
DOWNLOAD_BASE = 'http://download.tensorflow.org/models/object_detection/'

# Path to frozen detection graph. This is the actual model that is used for the object detection.
PATH_TO_CKPT = MODEL_NAME + '/frozen_inference_graph.pb'

# List of the strings that is used to add correct label for each box.

NUM_CLASSES = 90


# ## Download Model

# In[5]:
print( " Downloading model ")

opener = urllib.request.URLopener()
downloaded_model_aboslute_path = os.path.join(MODELS_HOME_DIR,MODEL_FILE)
opener.retrieve(DOWNLOAD_BASE + MODEL_FILE, downloaded_model_aboslute_path)
tar_file = tarfile.open(downloaded_model_aboslute_path)
for file in tar_file.getmembers():
  file_name = os.path.basename(file.name)
  if 'frozen_inference_graph.pb' in file_name:
    tar_file.extract(file, os.getcwd())

print (" Loading frozen model into memory")
# ## Load a (frozen) Tensorflow model into memory.

# In[6]:


detection_graph = tf.Graph()
with detection_graph.as_default():
  od_graph_def = tf.GraphDef()
  with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
    serialized_graph = fid.read()
    od_graph_def.ParseFromString(serialized_graph)
    tf.import_graph_def(od_graph_def, name='')


# ## Loading label map
# Label maps map indices to category names, so that when our convolution network predicts `5`, we know that this corresponds to `airplane`.  Here we use internal utility functions, but anything that returns a dictionary mapping integers to appropriate string labels would be fine

# In[7]:


label_map = label_map_util.load_labelmap(PATH_TO_LABELS)
categories = label_map_util.convert_label_map_to_categories(label_map, max_num_classes=NUM_CLASSES, use_display_name=True)
category_index = label_map_util.create_category_index(categories)


# ## Helper code

# In[8]:


def load_image_into_numpy_array(image):
  (im_width, im_height) = image.size
  return np.array(image.getdata()).reshape(
      (im_height, im_width, 3)).astype(np.uint8)


# # Detection

# In[9]:


# For the sake of simplicity we will use only 2 images:
# image1.jpg
# image2.jpg
# If you want to test the code with your images, just add path to the images to the TEST_IMAGE_PATHS.
TEST_IMAGE_PATHS = [ os.path.join(PATH_TO_TEST_IMAGES_DIR, '{}.jpg'.format(i)) for i in range(1, 5) ]

# Size, in inches, of the output images.
IMAGE_SIZE = (12, 8)


# In[11]:


with detection_graph.as_default():
  with tf.Session(graph=detection_graph) as sess:
    for image_path in TEST_IMAGE_PATHS:
      image = Image.open(image_path)
      # the array based representation of the image will be used later in order to prepare the
      # result image with boxes and labels on it.
      image_np = load_image_into_numpy_array(image)
      # Expand dimensions since the model expects images to have shape: [1, None, None, 3]
      image_np_expanded = np.expand_dims(image_np, axis=0)
      image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
      # Each box represents a part of the image where a particular object was detected.
      boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
      # Each score represent how level of confidence for each of the objects.
      # Score is shown on the result image, together with the class label.
      scores = detection_graph.get_tensor_by_name('detection_scores:0')
      classes = detection_graph.get_tensor_by_name('detection_classes:0')
      num_detections = detection_graph.get_tensor_by_name('num_detections:0')
      # Actual detection.
      (boxes, scores, classes, num_detections) = sess.run(
          [boxes, scores, classes, num_detections],
          feed_dict={image_tensor: image_np_expanded})
      # Visualization of the results of a detection.
      vis_util.visualize_boxes_and_labels_on_image_array(
          image_np,
          np.squeeze(boxes),
          np.squeeze(classes).astype(np.int32),
          np.squeeze(scores),
          category_index,
          use_normalized_coordinates=True,
          line_thickness=8)
      # plt.figure(figsize=IMAGE_SIZE)
      # plt.imshow(image_np)
      # i=0
      # plt.imsave(str(i)+'image',image_np )
# In[ ]:





# In[ ]: