Training Mask RCNN on Cloud TPU (TF 2.x)

---

Overview

This tutorial demonstrates how to run the Mask RCNN model using Cloud TPU with the COCO dataset.

Mask RCNN is a deep neural network designed to address object detection and image segmentation, one of the more difficult computer vision challenges.

The Mask RCNN model generates bounding boxes and segmentation masks for each instance of an object in the image. The model is based on the Feature Pyramid Network (FPN) and a ResNet50 backbone.

This tutorial uses the Tensorflow Keras API to train the model. The Keras API is a high-level TensorFlow API that can be used to build and run a machine learning model on Cloud TPU. The API simplifies the model development process by hiding most of the low-level implementation, which makes it easier to switch between TPU and other platforms such as GPU or CPU.

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

Warning: Mask RCNN uses a third-party dataset. Google provides no representation, warranty, or other guarantees about the validity or any other aspects of this dataset.

Objectives

• Prepare the COCO dataset
• Create a Cloud Storage bucket to hold your dataset and model output
• Set up TPU resources for training and evaluation
• Run training and evaluation on a single Cloud TPU or a Cloud TPU Pod

Costs

In this document, you use 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

Before starting this tutorial, check that your Google Cloud project is correctly set up.

1. Sign in to your Google Cloud account. If you're new to Google Cloud, create an account to evaluate how our products perform in real-world scenarios. New customers also get $300 in free credits to run, test, and deploy workloads.

2. In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

   Note: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.

   Go to project selector

3. Make sure that billing is enabled for your Google Cloud project.

4. In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

   Note: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.

   Go to project selector

5. Make sure that billing is enabled for your Google Cloud project.

This walkthrough uses billable components of Google Cloud. Check the Cloud TPU pricing page to estimate your costs.

Prepare the COCO dataset

This tutorial uses the COCO dataset. The dataset needs to be in TFRecord format on a Cloud Storage bucket to be used for the training.

If you already have the COCO dataset prepared on a Cloud Storage bucket that is located in the zone you will be using to train the model, you can go directly to single device training. Otherwise, use the following steps to prepare the dataset.

1. Open a Cloud Shell window.

   Open Cloud Shell

2. In your Cloud Shell, configure gcloud with your project ID.

   export PROJECT_ID=project-id
   gcloud config set project ${PROJECT_ID}
   

3. In your Cloud Shell, create a Cloud Storage bucket using the following command:

   Note: In the following command, replace bucket-name with the name you want to assign to your bucket.

   gcloud storage buckets create gs://bucket-name --project=${PROJECT_ID} --location=us-central2
   

4. Create a Compute Engine VM to download and preprocess the dataset. For more information, see Create and start a Compute Engine instance.

   $ gcloud compute instances create vm-name \
       --zone=us-central2-b \
       --image-family=ubuntu-2204-lts \
       --image-project=ubuntu-os-cloud \
       --machine-type=n1-standard-16 \
       --boot-disk-size=300GB
   

5. Connect to the Compute Engine VM using SSH:

   $ gcloud compute ssh vm-name --zone=us-central2-b
   

   When you connect to the VM, your shell prompt changes from username@projectname to username@vm-name.

6. Set up two variables, one for the storage bucket you created earlier and one for the directory that holds the training data (DATA_DIR) on the storage bucket.

   (vm)$ export STORAGE_BUCKET=gs://bucket-name
   

   (vm)$ export DATA_DIR=${STORAGE_BUCKET}/coco

7. Install the packages needed to pre-process the data.

   (vm)$ sudo apt-get update && \
     sudo apt-get install python3-pip && \
     sudo apt-get install -y python3-tk && \
     pip3 install --user Cython matplotlib opencv-python-headless pyyaml Pillow numpy absl-py tensorflow && \
     pip3 install --user "git+https://github.com/cocodataset/cocoapi#egg=pycocotools&subdirectory=PythonAPI"
   

8. Run the download_and_preprocess_coco.sh script to convert the COCO dataset into a set of TFRecord files (*.tfrecord) that the training application expects.

   (vm)$ git clone https://github.com/tensorflow/tpu.git
   (vm)$ sudo bash tpu/tools/datasets/download_and_preprocess_coco.sh ./data/dir/coco
   

   This installs the required libraries and then runs the preprocessing script. It outputs *.tfrecord files in your local data directory. The COCO download and conversion script takes approximately one hour to complete.

9. Copy the data to your Cloud Storage bucket.

   After you convert the data into the TFRecord format, copy the data from local storage to your Cloud Storage bucket using the gcloud CLI. You must also copy the annotation files. These files help validate the model's performance.

   (vm)$ gcloud storage cp ./data/dir/coco/*.tfrecord ${DATA_DIR}
   (vm)$ gcloud storage cp ./data/dir/coco/raw-data/annotations/*.json ${DATA_DIR}
   

10. Disconnect from the Compute Engine VM:

    (vm)$ exit
    

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

11. Delete your Compute Engine VM:

    $ gcloud compute instances delete vm-name \
    --zone=us-central2-b
    

Cloud TPU single device training

Note: Do you want to train the model using a Cloud TPU Pod? See TPU Pod training.

Important: Set up all resources in the same region and zone to reduce network latency and network costs.

1. Open a Cloud Shell window.

   Open Cloud Shell

2. Create an environment 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 the 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 Google Cloud API calls with your credentials.

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

   gcloud beta services identity create --service tpu.googleapis.com --project ${PROJECT_ID}
   

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

   [e-mail]
   

5. Export TPU setup variables

   Export your project ID, the name you want to use for your TPU resources, and the zone where you will train the model and store any training-related data.

   $ export TPU_NAME=mask-rcnn-tutorial
   $ export ZONE=europe-west4-a
   

6. Launch a Compute Engine VM and Cloud TPU.

   $ gcloud compute tpus tpu-vm create mask-rcnn-tutorial \
      --zone=${ZONE} \
      --accelerator-type=v3-8 \
      --version=tpu-vm-tf-2.18.0-pjrt
     

   Command flag descriptions

   zone

   The zone where you plan to create your Cloud TPU.

   accelerator-type

   The accelerator type specifies the version and size of the Cloud TPU you want to create. For more information about supported accelerator types for each TPU version, see TPU versions.

   version

   The Cloud TPU software version.

   Note: The first time you create a TPU VM on a project it takes several minutes to perform startup tasks such as SSH key propagation and API turnup.

   For more information on the gcloud command, see the gcloud Reference.

7. Connect to the Compute Engine instance using SSH. When you are connected to the VM, your shell prompt changes from username@projectname to username@vm-name:

   gcloud compute tpus tpu-vm ssh mask-rcnn-tutorial --zone=${ZONE}
   

   Key Point: From this point on, a prefix of (vm) $ means you should run the command on the Compute Engine VM instance.

8. Install TensorFlow requirements.

   (vm)$ pip3 install -r /usr/share/tpu/models/official/requirements.txt
   

9. Set the Cloud TPU name variable.

   (vm)$ export TPU_NAME=local
   

10. Set up the following environment variables, replacing bucket-name with the name of the Cloud Storage bucket that stores the COCO dataset:

    (vm)$ export STORAGE_BUCKET=gs://bucket-name
    

11. Add environment variables for the data and model directories.

    (vm)$ export DATA_DIR=${STORAGE_BUCKET}/coco
    (vm)$ export MODEL_DIR=${STORAGE_BUCKET}/mask-rcnn
    

12. Add some additional required environment variables:

    (vm)$ export RESNET_CHECKPOINT=gs://cloud-tpu-checkpoints/retinanet/resnet50-checkpoint-2018-02-07
    (vm)$ export TRAIN_FILE_PATTERN=${DATA_DIR}/train-*
    (vm)$ export EVAL_FILE_PATTERN=${DATA_DIR}/val-*
    (vm)$ export VAL_JSON_FILE=${DATA_DIR}/instances_val2017.json
    

13. Set the PYTHONPATH environment variable:

    (vm)$ export PYTHONPATH="${PYTHONPATH}:/usr/share/tpu/models"
    

14. When creating your TPU, if you set the --version parameter to a version ending with -pjrt, set the following environment variables to enable the PJRT runtime:

      (vm)$ export NEXT_PLUGGABLE_DEVICE_USE_C_API=true
      (vm)$ export TF_PLUGGABLE_DEVICE_LIBRARY_PATH=/lib/libtpu.so
    

15. Change to directory that stores the model:

    (vm)$ cd /usr/share/tpu/models/official/vision
    

The following script runs a sample training that trains for 10 training steps and 10 evaluation steps. It takes approximately 6 minutes to complete on a v3-8 TPU. To train to convergence takes about 22,500 steps and approximately 6 hours on a v3-8 TPU.

1. Run the following command to train Mask-RCNN model:

   (vm)$ python3 train.py \
     --tpu=${TPU_NAME} \
     --experiment=maskrcnn_resnetfpn_coco \
     --mode=train_and_eval \
     --config_file=configs/experiments/maskrcnn/r50fpn_640_coco_scratch_tpu4x4.yaml \
     --model_dir=${MODEL_DIR} \
     --params_override="task.train_data.input_path=${TRAIN_FILE_PATTERN},task.validation_data.input_path=${EVAL_FILE_PATTERN},task.annotation_file=${VAL_JSON_FILE},runtime.distribution_strategy=tpu,trainer.train_steps=10,trainer.validation_steps=10,task.train_data.global_batch_size=8,task.validation_data.global_batch_size=8"
   

   Command flag descriptions

   strategy_type

   The distribution strategy.

   tpu

   The name of your TPU.

   model_dir

   Specifies the directory where checkpoints and summaries are stored during model training. If the folder is missing, the program creates one. When using a Cloud TPU, the model_dir must be a Cloud Storage path (`gs://...`). You can reuse an existing folder to load current checkpoint data and to store additional checkpoints as long as the previous checkpoints were created using TPU of the same size and TensorFlow version.

   When the training completes, a message similar to the following appears:

   {'frcnn_box_loss': 0.033865165,
    'frcnn_cls_loss': 1.2535654,
    'learning_rate': 0.008266499,
    'mask_loss': 1.2039567,
    'model_loss': 2.821458,
    'rpn_box_loss': 0.034982488,
    'rpn_score_loss': 0.2950886,
    'total_loss': 4.340171,
    'training_loss': 4.340171}
   train | step:     10 | steps/sec:    0.1 | output:
   {'frcnn_box_loss': 0.033865165,
    'frcnn_cls_loss': 1.2535654,
    'learning_rate': 0.008266499,
    'mask_loss': 1.2039567,
    'model_loss': 2.821458,
    'rpn_box_loss': 0.034982488,
    'rpn_score_loss': 0.2950886,
    'total_loss': 4.340171,
    'training_loss': 4.340171}
   

   This is followed by output from the evaluation steps.

   You have now completed single-device training and evaluation. Use the following steps to delete the current single-device TPU resources.

2. Disconnect from the Compute Engine instance:

   (vm)$ exit
   

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

3. Delete the TPU resource.

   $ gcloud compute tpus tpu-vm delete mask-rcnn-tutorial \
       --zone=europe-west4-a
   

   Command flag descriptions

   zone

   The zone where your Cloud TPU resided.

At this point, you can either conclude this tutorial and clean up, or you can continue and explore running the model on Cloud TPU Pods.

Scaling your model with Cloud TPU Pods

Training your model on Cloud TPU Pods may require some changes to your training script. For information, see Training on TPU Pods.

TPU Pod training

Important: If you have already set up a Cloud TPU project, Service Account, storage bucket, and dataset for single device training, you can skip to setup and launch TPU resources.

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 Google Cloud API calls with your credentials.

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

   Service accounts allow the Cloud TPU service to access other Google Cloud services.

   gcloud beta services identity create --service tpu.googleapis.com --project $PROJECT_ID
   

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

   [e-mail]
   

5. If you previously prepared the COCO dataset and moved it to your storage bucket, you can use it again for Pod training. If you have not yet prepared the COCO dataset, prepare it now and return here to set up the training.

6. Launch a Cloud TPU Pod

   This tutorial specifies a v3-32 Pod. For other Pod options, see TPU versions.

   Note: If there is not enough capacity currently available to create the TPU Pod, you can queue your request using queued resources. Queued resources allow you to receive capacity once it becomes available. To request your Cloud TPU resources as queued resources, use the gcloud compute tpus queued-resources create command instead. For more information, see Manage Queued Resources.

   $ gcloud compute tpus tpu-vm create mask-rcnn-tutorial \
     --zone=${ZONE} \
     --accelerator-type=v3-32 \
     --version=tpu-vm-tf-2.18.0-pod-pjrt
   

   Command flag descriptions

   zone

   The zone where you plan to create your Cloud TPU.

   accelerator-type

   The accelerator type specifies the version and size of the Cloud TPU you want to create. For more information about supported accelerator types for each TPU version, see TPU versions.

   version

   The Cloud TPU software version.

   Note: The first time you run gcloud on a project it takes about 5 minutes to perform startup tasks such as SSH key propagation and API turnup.
   Note: The first time you run gcloud on a project it takes about 5 minutes to perform startup tasks such as SSH key propagation and API turnup.

7. Connect to the Compute Engine instance using SSH. When you are connected to the VM, your shell prompt changes from username@projectname to username@vm-name:

   gcloud compute tpus tpu-vm ssh mask-rcnn-tutorial --zone=${ZONE}
   

   Key Point: From this point on, a prefix of (vm) $ means you should run the command on the Compute Engine VM instance.

8. Install TensorFlow requirements.

   (vm)$ pip3 install -r /usr/share/tpu/models/official/requirements.txt
   

9. The training script requires an extra package. Install it now:

   (vm)$ pip3 install --user tensorflow-model-optimization>=0.1.3
   

10. Set the Cloud TPU name variable.

    (vm)$ export TPU_NAME=mask-rcnn-tutorial
    

11. Set up the following environment variables, replacing bucket-name with the name of your Cloud Storage bucket:

    (vm)$ export STORAGE_BUCKET=gs://bucket-name
    

12. Add some additional required environment variables:

    (vm)$ export RESNET_CHECKPOINT=gs://cloud-tpu-checkpoints/retinanet/resnet50-checkpoint-2018-02-07
    (vm)$ export DATA_DIR=${STORAGE_BUCKET}/coco
    (vm)$ export TRAIN_FILE_PATTERN=${DATA_DIR}/train-*
    (vm)$ export EVAL_FILE_PATTERN=${DATA_DIR}/val-*
    (vm)$ export VAL_JSON_FILE=${DATA_DIR}/instances_val2017.json
    (vm)$ export MODEL_DIR=${STORAGE_BUCKET}/mask-rcnn-pod
    

13. Set the PYTHONPATH environment variable:

    (vm)$ export PYTHONPATH="/usr/share/tpu/models:${PYTHONPATH}"
    (vm)$ export TPU_LOAD_LIBRARY=0
    

14. Change to directory that stores the model:

    (vm)$ cd /usr/share/tpu/models/official/vision
    

    (vm)$ cd /usr/share/models/official/legacy/detection

15. Train the model:

    This procedure trains the model on the COCO dataset for 10 training steps. This training takes approximately 10 minutes on a v3-32 Cloud TPU.

    Note: Once training starts, you can verify the training is running by monitoring your TPU activity in the Cloud TPU console. From Compute Engine -> TPUs, click your TPU entry. This will bring up a page with your TPU configuration details and a link to MONITORING. Click MONITORING to see the current TPU activity.

    (vm)$ python3 train.py \
      --tpu=${TPU_NAME} \
      --experiment=maskrcnn_resnetfpn_coco \
      --mode=train_and_eval \
      --config_file=configs/experiments/maskrcnn/r50fpn_640_coco_scratch_tpu4x4.yaml \
      --model_dir=${MODEL_DIR} \
      --params_override="task.train_data.input_path=${TRAIN_FILE_PATTERN},task.validation_data.input_path=${EVAL_FILE_PATTERN},task.annotation_file=${VAL_JSON_FILE},runtime.distribution_strategy=tpu,trainer.train_steps=10,trainer.validation_steps=10,task.train_data.global_batch_size=256,task.validation_data.global_batch_size=256"
    

    Command flag descriptions

    tpu

    The name of your TPU.

    model_dir

    Specifies the directory where checkpoints and summaries are stored during model training. If the folder is missing, the program creates one. When using a Cloud TPU, the model_dir must be a Cloud Storage path (gs://...). You can reuse an existing folder to load current checkpoint data and to store additional checkpoints as long as the previous checkpoints were created using Cloud TPU of the same size and TensorFlow version.

    params_override

    A JSON string that overrides default script parameters.

When the training completes, a message similar to the following appears:

 I0706 19:47:16.108213 139955064548416 controller.py:457] train | step: 10 | steps/sec:    0.1 | output:
    {'frcnn_box_loss': 0.05632668,
     'frcnn_cls_loss': 1.3012192,
     'learning_rate': 0.008266499,
     'mask_loss': 1.2371812,
     'model_loss': 2.9746659,
     'rpn_box_loss': 0.08227444,
     'rpn_score_loss': 0.2976642,
     'total_loss': 4.493513,
     'training_loss': 4.493513}
train | step:     10 | steps/sec:    0.1 | output:
    {'frcnn_box_loss': 0.05632668,
     'frcnn_cls_loss': 1.3012192,
     'learning_rate': 0.008266499,
     'mask_loss': 1.2371812,
     'model_loss': 2.9746659,
     'rpn_box_loss': 0.08227444,
     'rpn_score_loss': 0.2976642,
     'total_loss': 4.493513,
     'training_loss': 4.493513}
 

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.

After running the training, delete the TPU VM and remove your storage bucket.

1. Disconnect from the Compute Engine instance, if you have not already done so:

   (vm)$ exit
   

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

2. Delete your Cloud TPU and Compute Engine resources.

   $ gcloud compute tpus tpu-vm delete mask-rcnn-tutorial \
   --zone=${ZONE}
   

3. Verify the resources have been deleted by running gcloud compute tpus tpu-vm list. The deletion might take several minutes. The output from the following command shouldn't include any of the TPU resources created in this tutorial:

   $ gcloud compute tpus tpu-vm list --zone=${ZONE}
   

4. Run the gcloud CLI as shown, replacing bucket-name with the name of the Cloud Storage bucket you created for this tutorial:

   $ gcloud storage rm gs://bucket-name --recursive
   

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 dataset. 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 and Tune hyperparameters.

Inference

Once you have trained your model, you can use it for inference (also called prediction). You can use the Cloud TPU inference converter tool to prepare and optimize a TensorFlow model for inference on Cloud TPU v5e. For more information about inference on Cloud TPU v5e, see Cloud TPU v5e inference introduction.

Explore the TPU tools in TensorBoard.