Name: Predictive-Industrial-Visual-Analysis
Owner: International Business Machines
Description: Predictive Industrial Visual Analysis using Watson Visual Recognition, IBM Cloud Functions and Cloudant database
Created: 2017-10-23 19:25:42.0
Updated: 2018-04-27 18:02:47.0
Pushed: 2018-04-27 18:02:47.0
Homepage: http://industrial-visual-analysis.mybluemix.net/
Size: 32438
Language: JavaScript
GitHub Committers
User | Most Recent Commit | # Commits |
---|
Other Committers
User | Most Recent Commit | # Commits |
---|
Skill Level: Beginner
N.B: All services used in this repo are Lite plans.
In this code pattern, we will identify industrial equipment for various damages upon visual inspection by using machine learning classification techniques. Using Watson Visual Recognition, we will analyze the image against a trained classifier to inspect oil and gas pipelines with six identifiers - Normal, Burst, Corrosion, Damaged Coating, Joint Failure and Leak. For each image we will provide a percent match with each of the category, on how closely the image matches one of the damaged identifiers or the Normal identifier. This data can then be used to create a dashboard to the pipelines needing immediate attention to no attention.
The images data is stored in a Cloudant database which makes it easier to connect remote devices (including drones) to capture images. The database can store different properties of the images like location and description. This code pattern demonstrates IBM Cloud Functions (OpenWhisk) to trigger microservice as an image is added to the Cloudant database. The microservice performs the Visual Recognition analysis and updates the Cloudant database with the analysis data.
When the reader has completed this code pattern, they will understand how to:
Follow these steps to setup and run the application. The steps are described in detail below.
Create the Watson Visual Recognition service in IBM Cloud. You will need the API Key
.
Open a command line interface (CLI) on your desktop and clone this repo:
clone https://github.com/IBM/Predictive-Industrial-Visual-Analysis
Go to the folder where the images are placed
redictive-Industrial-Visual-Analysis/vr-image-data
Here we will create a classifier using the zipped images to train the Watson Visual-Recognition service. The images in each zipped folder are used to make the Watson VR service become familiar with the images that relate to the different categories (Corrosion, Leak, etc.). Run the following command to submit all 6 sets of images to the Watson service classifier:
-X POST -u "apikey:{INSERT-YOUR-IAM-APIKEY-HERE}" -F "Bursted_Pipe_positive_examples=@Burst_Images.zip" -F "Corroded_Pipe_positive_examples=@Corrosion_Images.zip" -F "Damaged_Coating_positive_examples=@Damaged_Coating_Images.zip" -F "Joint_Failure_positive_examples=@Joint_Failure_Images.zip" -F "Pipe_Leak_positive_examples=@Leak_Images.zip" -F "Normal_Condition_positive_examples=@Normal_Condition.zip" -F "name=OilPipeCondition" "https://gateway.watsonplatform.net/visual-recognition/api/v3/classifiers?version=2018-03-19"
The response from above will provide you with a status on the submission and will give you a CLASSIFIER_ID
. Please copy this for future use as well. After executing the above command, you can view the status of your Watson service and whether it has finished training on the images you submitted. You can check the status like this:
-X GET -u "apikey:{INSERT-YOUR-IAM-APIKEY-HERE}" "https://gateway.watsonplatform.net/visual-recognition/api/v3/classifiers/{INSERT-CLASSIFIER-ID-HERE}?api_key={INSERT-API-KEY-HERE}&version=2018-03-19"
You can find more information on working with your classifier here
Create the Cloudant NoSQL service in IBM Cloud.
Create a new database in Cloudant called image_db
Next, create a view on the database with the design name image_db_images
, index name image_db.images
, and use the following map function:
tion (doc) {
doc.type == 'image_db.image' ) {
it(doc);
We will now set up the IBM Cloud Functions (OpenWhisk) using Bluemix CLI.
Install the Cloud Functions Plugin
lugin install Cloud-Functions -r Bluemix
Log in to IBM Cloud, and target a Region (i.e api.ng.bluemix.net), Organization (i.e Raheel.Zubairy) and Space (i.e dev).
ogin -a {INSERT REGION} -o {INSERT ORGANIZATION} -s {INSERT SPACE}
We will need the API authentication key and host.
Command to retrieve API host:
sk property get --apihost
Command to retrieve API authentication key:
sk property get --auth
N.B: make sure what plan (Lite, etc.) you are associating when creating this service.
You will need to provide credentials to your Cloudant NoSQL database and Watson Visual Recognition service, and Cloud Functions Host/Auth information retrieved in the previous step, into a .env file
. Copy the sample .env.example
file using the following command:
env.example .env
and fill in your credentials and your VR Classifier name.
m cloudant NoSQL database
DANT_USERNAME=
DANT_PASSWORD=
DANT_HOST=
DANT_URL=
DANT_DB=image_db
m Watson Visual Recognition Service
EY=
RL=
LASSIFIERS=default,OilPipeCondition_1063693116
m OpenWhisk Functions Service in IBM Cloud
TIONS_APIHOST=
TIONS_AUTHORIZATION=
We will now run the setup_functions.sh
file to set up the microservice which triggers the Visual Recognition analysis as an image is added to the Cloudant database.
d +x setup_functions.sh
tup_functions.sh --install
The above command will setup the OpenWhisk actions for you, there should be no need to do anything else if you see an Install Complete message with green OK signs in the CLI.
In IBM Cloud, look for Functions
in Catalog
There you will see a UI to Manage
and Monitor
the service. In addition, it has information for Getting Started
and even Develop
actions.
To run the app, go to the `Industrial-Visual-Analysis
` folder and run the following commands.
install
start
Test your application by going to: http://localhost:3000/
You can push the app to IBM Cloud by first editing the `manifest file
` file and then using cloud foundry cli commands.
Edit the manifest.yml
file in the folder that contains your code and replace with a unique name for your application. The name that you specify determines the application's URL, such as your-application-name.mybluemix.net
. Additionally - update the service names so they match what you have in IBM Cloud. The relevant portion of the manifest.yml
file looks like the following:
ications:
th: .
mory: 256M
stances: 1
main: mybluemix.net
me: {industrial-visual-analysis}
sk_quota: 1024M
rvices:
{cloudant}
{visual-recognition}
In the command line use the following command to push the application to IBM Cloud:
pp push YOUR_APP_NAME
The app has the following functions:
The homepage displays a quick dashboard showing the number of images in the Cloudant database and how many of them have Watson VR analysis completed. It will also provide a count of how many images were deemed as “Needing attention” based on the response the Watson service provided when classifying the images.
You have the ability to see all the images in one single page.
Click on each image to pull up a detailed page providing information on one single event (image). You will be able to see information on what the Watson Visual Recognition service saw in the image and the confidence levels. You can continue to train the service by using the thumbs up and thumbs down next to each percent match.
You can click the Upload New Image
button to send images to the Cloudant database. There are sample images in the sample-images
folder to try out.
This code pattern can be extended by adding a Drone to take images. A DJI drone can be used to capture images and configured to send images to our Cloudant database. As the image is received by the Cloudant database, the VR analysis and image detail can be displayed through the web UI.
If you invoke GET /classifiers
with verbose=1
what do you see? If that list is empty, and you get this error message, you should open an IBM Cloud support ticket. If it's not empty, you should use DELETE /classifiers/{classifier_id}
to remove the existing classifier so that you can create your new one.
The setup_functions.sh
have different commands to uninstall, re-install or update IBM Cloud Functions. And to view the env credentials used by IBM Cloud Functions.
Uninstall:
tup_functions.sh --uninstall
Re-install:
tup_functions.sh --reinstall
Show env cred:
tup_functions.sh --env
Update:
tup_functions.sh --update
To troubleshoot your IBM Cloud application, use the logs. To see the logs, run:
pp logs <application-name> --recent