Running Deeplab-v3 on Cloud TPU

This tutorial shows you how to train the Deeplab-v3 model on Cloud TPU.

The instructions below assume you are already familiar with running a model on Cloud TPU. If you are new to Cloud TPU, you can refer to the Quickstart for a basic introduction.

If you plan to train on a TPU Pod slice, review Training on TPU Pods to understand parameter changes required for Pod slices.

This model is an image semantic segmentation model. Image semantic segmentation models focus on identifying and localizing multiple objects in a single image. This type of model is frequently used in machine learning applications such as autonomous driving, geospatial image processing, and medical imaging.

In this tutorial, you'll run a training model against the PASCAL VOC 2012 dataset. For more information on this data set, see The PASCAL Visual Object Classes Homepage.


  • Create a Cloud Storage bucket to hold your dataset and model output.
  • Install the required packages.
  • Download and convert the PASCAL VOC 2012 dataset.
  • Train the Deeplab model.
  • Evaluate the Deeplab model.


This tutorial uses the following billable components of Google Cloud:

  • Compute Engine
  • Cloud TPU
  • Cloud Storage

To generate a cost estimate based on your projected usage, use the pricing calculator. New Google Cloud users might be eligible for a free trial.

Before you begin

This section provides information on setting up Cloud Storage bucket and a Compute Engine VM.

  1. Open a Cloud Shell window.

    Open Cloud Shell

  2. Create a variable for your project's ID.

    export PROJECT_ID=project-id
  3. Configure Google Cloud CLI to use the project where you want to create Cloud TPU.

    gcloud config set project ${PROJECT_ID}

    The first time you run this command in a new Cloud Shell VM, an Authorize Cloud Shell page is displayed. Click Authorize at the bottom of the page to allow gcloud to make GCP API calls with your credentials.

  4. Create a Service Account for the Cloud TPU project.

    gcloud beta services identity create --service --project $PROJECT_ID

    The command returns a Cloud TPU Service Account with following format:

  5. Create a Cloud Storage bucket using the following command:

    gsutil mb -p ${PROJECT_ID} -c standard -l us-central1 -b on gs://bucket-name

    This Cloud Storage bucket stores the data you use to train your model and the training results.

    In order for the Cloud TPU to read and write to the storage bucket, the Service Account for your project needs read/write or Admin permissions on it. See the section on storage buckets for how to view and set those permissions.

  6. Launch a Compute Engine VM using the ctpu up command.

    $ ctpu up --project=${PROJECT_ID} \
     --zone=us-central1-b \
     --machine-type=n1-standard-8 \
     --vm-only \
     --tf-version=1.15.5 \

    Command flag descriptions

    Your GCP project ID
    The zone where you plan to create your Cloud TPU.
    The machine type of the Compute Engine VM to create.
    Create a VM only. By default the ctpu up command creates a VM and a Cloud TPU.
    The version of Tensorflow ctpu installs on the VM.
    The name of the Cloud TPU to create.
  7. The configuration you specified appears. Enter y to approve or n to cancel.

  8. When the ctpu up command has finished executing, verify that your shell prompt has changed from username@projectname to username@vm-name. This change shows that you are now logged into your Compute Engine VM.

    gcloud compute ssh deeplab-tutorial --zone=us-central1-b

As you continue these instructions, run each command that begins with (vm)$ in your VM session window.

Install additional packages

For this model, you need to install the following additional packages on your Compute Engine instance:

  • jupyter
  • matplotlib
  • PrettyTable
  • tf_slim
  (vm)$ pip3 install --user jupyter
  (vm)$ pip3 install --user matplotlib
  (vm)$ pip3 install --user PrettyTable
  (vm)$ pip3 install --user tf_slim
  1. Create environment variables for your storage bucket and TPU name.

    (vm)$ export STORAGE_BUCKET=gs://bucket-name
    (vm)$ export TPU_NAME=deeplab-tutorial
    (vm)$ export DATA_DIR=${STORAGE_BUCKET}/deeplab_data
    (vm)$ export MODEL_DIR=${STORAGE_BUCKET}/deeplab_model
    (vm)$ export PYTHONPATH=${PYTHONPATH}:/usr/share/models/research:/usr/share/models/research/slim

Prepare the data set

  1. Download and convert the PASCAL VOC 2012 dataset

    This model uses the PASCAL VOC 2012 dataset for training and evaluation. Run the following script to download the dataset and convert it to TensorFlow's TFRecord format:

     (vm)$ bash /usr/share/models/research/deeplab/datasets/
  2. Download the pretrained checkpoint

    In this step, you download the modified resnet 101 pretrained checkpoint. To start, download the checkpoint:

     (vm)$ wget

    Then, extract the contents of the tar file:

     (vm)$ tar -vxf resnet_v1_101_2018_05_04.tar.gz
  3. Upload data to your Cloud Storage bucket

    At this point, you can now upload your data to the Cloud Storage bucket you created earlier:

    (vm)$ gsutil -m cp -r pascal_voc_seg/tfrecord ${DATA_DIR}/tfrecord
    (vm)$ gsutil -m cp -r resnet_v1_101 ${DATA_DIR}

Create Cloud TPU resource

Run the following command to create your Cloud TPU.

  (vm)$ ctpu up --project=${PROJECT_ID} \
  --tpu-only \
  --tf-version=1.15.5 \
  --tpu-size=v3-8 \

Train the model

Run the training script for 2000 training steps. This will take approximately 20 minutes. To run to convergence, remove the --train_steps=2000 flag from the training script command line. Running to convergence takes about 10 hours.

(vm)$ python3 /usr/share/tpu/models/experimental/deeplab/ \
--mode='train' \
--num_shards=8 \
--alsologtostderr=true \
--model_dir=${MODEL_DIR} \
--dataset_dir=${DATA_DIR}/tfrecord \
--init_checkpoint=${DATA_DIR}/resnet_v1_101/model.ckpt \
--model_variant=resnet_v1_101_beta \
--image_pyramid=1. \
--aspp_with_separable_conv=false \
--multi_grid=1 \
--multi_grid=2 \
--multi_grid=4 \
--decoder_use_separable_conv=false \
--train_split='train' \
--train_steps=2000 \

Evaluating the model on a Cloud TPU device.

When the training completes, you can evaluate the model. To do so, change the --mode flag from train to eval:

(vm)$ python3 /usr/share/tpu/models/experimental/deeplab/ \
--mode='eval' \
--num_shards=8 \
--alsologtostderr=true \
--model_dir=${MODEL_DIR} \
--dataset_dir=${DATA_DIR}/tfrecord \
--init_checkpoint=${DATA_DIR}/resnet_v1_101/model.ckpt \
--model_variant=resnet_v1_101_beta \
--image_pyramid=1. \
--aspp_with_separable_conv=false \
--multi_grid=1 \
--multi_grid=2 \
--multi_grid=4 \
--decoder_use_separable_conv=false \
--train_split='train' \
--tpu=${TPU_NAME} \

Clean up

To avoid incurring charges to your Google Cloud account for the resources used in this tutorial, either delete the project that contains the resources, or keep the project and delete the individual resources.

  1. Disconnect from the Compute Engine VM:

    (vm)$ exit

    Your prompt should now be username@projectname, showing you are in the Cloud Shell.

  2. In your Cloud Shell, run ctpu delete with the --zone flag you used when you set up the Cloud TPU to delete your Compute Engine VM and your Cloud TPU:

    $ ctpu delete --project=${PROJECT_ID} \
      --zone=us-central1-b \
  3. Run ctpu status to make sure you have no instances allocated to avoid unnecessary charges for TPU usage. The deletion might take several minutes. A response like the one below indicates there are no more allocated instances:

    $ ctpu status --project=${PROJECT_ID} \
      --name=deeplab-tutorial \
    2018/04/28 16:16:23 WARNING: Setting zone to "us-central1-b"
    No instances currently exist.
            Compute Engine VM:     --
            Cloud TPU:             --
  4. Run gsutil as shown, replacing bucket-name with the name of the Cloud Storage bucket you created for this tutorial:

    $ gsutil rm -r gs://bucket-name

What's next

The TensorFlow Cloud TPU tutorials generally train the model using a sample dataset. The results of this training are not usable for inference. To use a model for inference, you can train the data on a publicly available dataset or your own data set. TensorFlow models trained on Cloud TPUs generally require datasets to be in TFRecord format.

You can use the dataset conversion tool sample to convert an image classification dataset into TFRecord format. If you are not using an image classification model, you will have to convert your dataset to TFRecord format yourself. For more information, see TFRecord and tf.Example.

Hyperparameter tuning

To improve the model's performance with your dataset, you can tune the model's hyperparameters. You can find information about hyperparameters common to all TPU supported models on GitHub. Information about model-specific hyperparameters can be found in the source code for each model. For more information on hyperparameter tuning, see Overview of hyperparameter tuning, Using the Hyperparameter tuning service, and Tune hyperparameters.


Once you have trained your model you can use it for inference (also called prediction). AI Platform is a cloud-based solution for developing, training, and deploying machine learning models. Once a model is deployed, you can use the AI Platform Prediction service.