NGC | Catalog
CatalogHelm ChartsIBM Visual Insights

IBM Visual Insights

Logo for IBM Visual Insights
IBM Visual Insights makes computer vision with deep learning more accessible to business users. The software includes an intuitive toolset that empowers subject matter experts to label, train, and deploy deep learning.
IBM Corporation
Latest Version
Compressed Size
703.27 KB
April 4, 2023


By pulling, downloading and/or using the program you agree to be bound by the end-user license at the following link:

It is required to register with IBM before pulling, downloading and/or using the program. Link to register.

What is IBM Visual Insights Evaluation Edition

IBM Visual Insights makes computer vision with deep learning more accessible to business users. The software includes an intuitive toolset that empowers subject matter experts to label, train, and deploy deep learning vision models, without coding or deep learning expertise.

Designed for business users, the platform provides auto labelling and data augmenting capabilities to accelerate labelling of volumes of images & videos and even work with limited datasets. Bundled with several deep learning frameworks, models for training classifiers, object detectors, image segmentation and Action recognition are core features of the software. Once the models are trained, IBM Visual Insights lets you export the models to iOS devices, NVIDIA TX2/Nano and even to the cloud of your choice.

System Requirements

Before running the containers for IBM Visual Insights Evaluation Edition, please ensure that your system meets the following HW and SW requirements.

Software requirements

  • Red Hat Enterprise Linux® (RHEL) RHEL 7.6 ALT (little endian) for POWER9™
  • RHEL 7.7 for x86
  • Ubuntu 18.04

Hardware requirements

  • POWER8 S822LC (8335-GTB) or POWER9 AC922 with at least one NVIDIA NVLink capable GPU
  • POWER9 IC922 with at least one NVIDIA T4 GPU
  • x86 system with at least one NVIDIA Pascal, Volta, or Turing-architecture GPU
  • 64 GB of memory
  • Ethernet network interface
  • 75 GB of storage for installation, and at least 40 GB of storage for runtime. See Disk space requirements for details.

Installation Instructions for NGC Clients

  1. Ensure Helm 2 / Tiller is available in your cluster.

  2. Ensure an Ingress is configured for your cluster.

  3. Create a namespace for the application. This is optional, but recommended. Subsequent commands will use CustomNamespace namespace as an example. Replace this with your own custom namespace.

    kubectl create namespace CustomNamespace
  4. Create an appropriate ClusterRoleBinding to enable PowerAI Vision to query Kubernetes. To create this, copy the below text into a crb.yaml file, where CustomNamespace is your custom namespace name:

    kind: ClusterRoleBinding
      name: CustomNamespace-crb
      kind: ClusterRole
      name: cluster-admin
        kind: ServiceAccount
        name: default
        namespace: CustomNamespace
  5. Apply the ClusterRoleBinding:

    kubectl create -f crb.yaml
  6. Verify that you have a minimum of 40 GB of persistent storage. If your kubernetes cluster has dynamic provisioned storage, you can use it for your persistent storage. To manually create persistent volumes, see the Kubernetes Storage Topics to learn how to work with persistent volumes. After you create the persistent volume, you must make the volume sharable across all nodes in the cluster.

  7. IMPORTANT: Containers within IBM Visual Insights run as non-root users for security reasons. You must set appropriate permissions in order for the application to start up correctly.

    mkdir -p ${VOL_DIR}/data ${VOL_DIR}/run/logstash ${VOL_DIR}/run/elasticsearch ${VOL_DIR}/run/mongodb ${VOL_DIR}/run/pgsql
    chown ${AIV_USER}:${AIV_USER} ${VOL_DIR} ${VOL_DIR}/run
    chown -R ${AIV_USER}:${AIV_USER} ${VOL_DIR}/data
    chown 1000:1000 ${VOL_DIR}/run/logstash ${VOL_DIR}/run/elasticsearch
    chown 999 ${VOL_DIR}/run/mongodb
    chown 999 ${VOL_DIR}/run/pgsql
  8. Create an appropriate imagePullSecret to allow IBM Visual Insights to dynamically pull container images from NGC as needed. Be sure to associate the secret with the namespace you created earlier. Refer to the Kubernetes documentation for details about how to pull an image from a private registry.

    #Note: This is a simple example. Your cluster may be different. Please note the '$oauthtoken' below. This is to provide NGC with the literal '$oauthtoken' username beginning with a dollar sign ($) expected to log in.
    kubectl create secret docker-registry --namespace CustomNamespace ngc --docker-username='$oauthtoken' --docker-password=YOUR-NGC-TOKEN
    kubectl patch serviceaccount --namespace CustomNamespace default -p '{"imagePullSecrets": [{"name": "ngc"}]}'
    #   where: YOUR-NGC-TOKEN  is your API token to access the NGC private registry
  9. Fetch the helm chart per the helm pull command provided by NGC above.

  10. Deploy the chart with variables set based on your processor architecture (POWER9 or x86):

    For x86 architecture with NVIDIA GPUs (amd64 or x86_64):
      helm install --name trial --namespace CustomNamespace ./ibm-visual-insights-prod-2.0.0.tgz --set image.pullPolicy=IfNotPresent --set image.releaseTag= --set
    For IBM Power architecture with NVIDIA GPUs (ppc64el or ppc64le or POWER):
      helm install --name trial --namespace CustomNamespace ./ibm-visual-insights-prod-2.0.0.tgz --set image.pullPolicy=IfNotPresent --set image.releaseTag= --set
  11. Wait for all pods to become ready, then log in via your cluster ingress. By default, the chart will create an ingress-controller that follows the visual-insights-HELMRELEASENAME pattern (e.g. /visual-insights-trial).

    helm status trial
  12. Log in to the application at https://ingress.hostname/visual-insights-trial . IMPORTANT: If you changed the chart deployment name from trial to another value , log in at https://ingress.hostname/visual-insights-HELMRELEASENAME The default username/password admin / passw0rd.

  13. Change the default admin password. Note that you may not create new users in the Trial edition of this software.

    kubectl run --rm -i --restart=Never usermgt --namespace CustomNamespace  -- modify  --user admin --password INSERT-PASSWORD-HERE --release trial

Reach us

Ask us your questions on our forum.

Next Steps

Suggested Reading