Creating and containerizing a TensorFlow Go application

In an earlier post, I discussed using a TensorFlow model from a Go application. With the TensorFlow bindings for Go, you can load a model that was exported with TensorFlow’s SavedModelBuilder module. That module saves a “snapshot” of a trained model which can be used for inference.

In this post, we will actually use the model in a web application. The application presents the user with a page to upload an image:

The upload page

The class and its probability is displayed, including the processed image:

Clearly a hen!

The source code of the application can be found at https://github.com/gbaeke/nasnet-go. If you just want to try the application, use Docker and issue the following command (replace port 80 with another port if there is a conflict):

docker run -p 80:9090 -d gbaeke/nasnet

The image is around 2.55GB in size so be patient when you first run the application. When the container has started, open your browser at http://localhost to see the upload page.

To quickly try it, you can run the container on Azure Container Instances. If you use the Portal, specify port 9090 as the container port.

Nasnet container in ACI

A closer look at the appN

**UPDATE**: since first publication, the http handler code was moved into from main.go to handlers/handlers.go

In the init() function, the nasnet model is loaded with tf.LoadSavedModel. The ImageNet categories are also loaded with a call to getCategories() and stored in categories which is a map of int to a string array.

In main(), we simply print the TensorFlow version (1.12). Next, http.HandleFunc is used to setup a handler (upload func) when users connect to the root of the web app.

Naturally, most of the logic is in the upload function. In summary, it does the following:

  • when users just navigate to the page (HTTP GET verb), render the upload.gtpl template; that template contains the upload form and uses a bit of bootstrap to make it just a bit better looking (and that’s already an overstatement); to learn more about Go web templates, see this link.
  • when users submit a file (POST), the following happens:
    • read the image
    • convert the image to a tensor with the getTensor function; getTensor returns a *tf.Tensor; the tensor is created from a [1][224][224][3] array; note that each pixel value gets normalized by subtracting by 127.5 and then dividing by 127.5 which is the same preprocessing applied as in Keras (divide by 127.5 and subtract 1)
    • run a session by inputting the tensor and getting the categories and probabilities as output
    • look for the highest probability and save it, together with the category name in a variable of type ResultPageData (a struct)
    • the struct data is used as input for the response.gtpl template

Note that the image is also shown in the output. The processed image (resized to 224×224) gets converted to a base64-encoded string. That string can be used in HTML image rendering as follows (where {{.Picture}} in the template will be replaced by the encoded string):

 <img src="data:image/jpg;base64,{{.Picture}}"> 

Note that the application lacks sufficient error checking to gracefully handle the upload of non-image files. Maybe I’ll add that later! ūüėČ

Containerization

To containerize the application, I used the Dockerfile from https://github.com/tinrab/go-tensorflow-image-recognition but removed the step that downloads the InceptionV3 model. My application contains a ready to use NasnetMobile model.

The container image is based on tensorflow/tensorflow:1.12.0. It is further modified as required with the TensorFlow C API and the installation of Go. As discussed earlier, I uploaded a working image on Docker Hub.

Conclusion

Once you know how to use TensorFlow models from Go applications, it is easy to embed them in any application, from command-line tools to APIs to web applications. Although this application does server-side processing, you can also use a model directly in the browser with TensorFlow.js or ONNX.js. For ONNX, try https://microsoft.github.io/onnxjs-demo/#/resnet50 to perform image classification with ResNet50 in the browser. You will notice that it will take a while to get started due to the model being downloaded. Once the model is downloaded, you can start classifying images. Personally, I prefer the server-side approach but it all depends on the scenario.

Using TensorFlow models in Go

Image via www.vpnsrus.com

In earlier posts, I discussed hosting a deep learning model such as Resnet50 on Kubernetes or Azure Container Instances. The model can then be used as any API which receives input as JSON and returns a result as JSON.

Naturally, you can also run the model in offline scenarios and directly from your code. In this post, I will take a look at calling a TensorFlow model from Go. If you want to follow along, you will need Linux or MacOS because the Go module does not support Windows.

Getting Ready

I installed an Ubuntu Data Science Virtual Machine on Azure and connected to it with X2Go:

Data Science Virtual Machine (Ubuntu) with X2Go

The virtual machine has all the required machine learning tools installed such as TensorFlow and Python. It also has Visual Studio Code. There are some extra requirements though:

  • Go: follow the instructions here to download and install Go
  • TensorFlow C API: follow the instructions here to download and install the C API; the TensorFlow package for Go requires this; it is recommended to also build and run the Hello from TensorFlow C program to verify that the library works (near the bottom of the instructions page)

After installing Go and the TensorFlow C API, install the TensorFlow Go package with the following command:

go get github.com/tensorflow/tensorflow/tensorflow/go

Test the package with go test:

go test github.com/tensorflow/tensorflow/tensorflow/go

The above command should return:

ok      github.com/tensorflow/tensorflow/tensorflow/go  0.104s

The go get command installed the package in $HOME/go/src/github.com if you did not specify a custom $GOPATH (see this wiki page for more info).

Getting a model

A model describes how the input (e.g. an image for image classification) gets translated to an output (e.g. a list of classes with probabilities). The model contains thousands or even millions of parameters which means a model can be quite large. In this example, we will use NASNetMobile which can be used to classify images.

Now we need some code to save the model in TensorFlow format so that it can be used from a Go program. The code below is based on the sample code on the NASNetMobile page from modeldepot.io. It also does a quick test inference on a cat image.

import keras
from keras.applications.nasnet import NASNetMobile
from keras.preprocessing import image
from keras.applications.xception import preprocess_input, decode_predictions
import numpy as np
import tensorflow as tf
from keras import backend as K

sess = tf.Session()
K.set_session(sess)

model = NASNetMobile(weights="imagenet")
img = image.load_img('cat.jpg', target_size=(224,224))
img_arr = np.expand_dims(image.img_to_array(img), axis=0)
x = preprocess_input(img_arr)
preds = model.predict(x)
print('Prediction:', decode_predictions(preds, top=5)[0])

#save the model for use with TensorFlow
builder = tf.saved_model.builder.SavedModelBuilder("nasnet")

#Tag the model, required for Go
builder.add_meta_graph_and_variables(sess, ["atag"])
builder.save()
sess.close()

On the Ubuntu Data Science Virtual Machine, the above code should execute without any issues because all Python packages are already installed. I used the py35 conda environment. Use activate py35 to make sure you are in that environment.

The above code results in a nasnet folder, which contains the saved_model.pb file for the graph structure. The actual weights are in the variables subfolder. In total, the nasnet folder is around 38MB.

Great! Now we need a way to use the model from our Go program.

Using the saved model from Go

The model can be loaded with the LoadSavedModel function of the TensorFlow package. That package is imported like so:

import (
tf "github.com/tensorflow/tensorflow/tensorflow/go"
)

LoadSavedModel is used like so:

model, err := tf.LoadSavedModel("nasnet",
[]string{"atag"}, nil)
if err != nil {
log.Fatal(err)
}

The above code simply tries to load the model from the nasnet folder. We also need to specify the tag.

Next, we need to load an image and convert the image to a tensor with the following dimensions [1][224][224][3]. This is similar to my earlier ResNet50 post.

Now we need to pass the tensor to the model as input, and retrieve the class predictions as output. The following code achieves this:

output, err := model.Session.Run(
map[tf.Output]*tf.Tensor{
model.Graph.Operation("input_1").Output(0): input,
},
[]tf.Output{
model.Graph.Operation("predictions/Softmax").Output(0),
},
nil,
)
if err != nil {
log.Fatal(err)
}

What the heck is this? The run method is defined as follows:

func (s *Session) Run(feeds map[Output]*Tensor, fetches []Output, targets []*Operation) ([]*Tensor, error)

When you build a model, you can give names to tensors and operations. In this case the input tensor (of dimensions [1][224][224][3]) is called input_1 and needs to be specified as a map. The inference operation is called predictions/Softmax and the output needs to be specified as an array.

The actual predictions can be retrieved from the output variable:

predictions, ok := output[0].Value().([][]float32)
if !ok {
log.Fatal(fmt.Sprintf("output has unexpected type %T", output[0].Value()))
}

If you are not very familiar with Go, the code above uses type¬†assertion to verify that predictions is a 2-dimensional array of float32. If the type assertion succeeds, the predictions variable will contain the actual predictions: [[<probability class 1 (tench)>, <probability class 2 (goldfish)>, …]]

You can now simply find the top prediction(s) in the array and match them with the list of classes for NASNet (actually the ImageNet classes). I get the following output with a cat image:

Yep, it’s a tabby!

If you are wondering what image I used:

Tabby?

Conclusion

With Go’s TensorFlow bindings, you can load TensorFlow models from disk and use them for inference locally, without having to call a remote API. We used Python to prepare the model with some help from Keras.

Microsoft Face API with a local container

A few days ago, I obtained access to the Face container. It provides access to the Face API via a container you can run where you want: on your pc, at the network edge or in your datacenter. You should allocate 6 GB or RAM and 2 cores for the container to run well. Note that you still need to create a Face API resource in the Azure Portal. The container needs to be associated with the Azure Face API via the endpoint and access key:

Face API with a West Europe (Amsterdam) endpoint

I used the Standard tier, which charges 0.84 euros per 1000 calls. As noted, the container will not function without associating it with an Azure Face API resource.

When you gain access to the container registry, you can pull the container:

docker pull containerpreview.azurecr.io/microsoft/cognitive-services-face:latest

After that, you can run the container as follows (for API billing endpoint in West Europe):

docker run --rm -it -p 5000:5000 --memory 6g --cpus 2 containerpreview.azurecr.io/microsoft/cognitive-services-face Eula=accept Billing=https://westeurope.api.cognitive.microsoft.com/face/v1.0 ApiKey=YOUR_API_KEY

The container will start. You will see the output (–it):

Running Face API container

And here’s the spec:

API spec Face API v1

Before showing how to use the detection feature, note that the container needs Internet access for billing purposes. You will not be able to run the container in fully offline scenarios.

Over at https://github.com/gbaeke/msface-go, you can find a simple example in Go that uses the container. The Face API can take a byte stream of an image or a URL to an image. The example takes the first approach and loads an image from disk as specified by the -image parameter. The resulting io.Reader is passed to the getFace function which does the actual call to the API (uri = http://localhost:5000/face/v1.0/detect):

request, err := http.NewRequest("POST", uri+"?returnFaceAttributes="+params, m)
request.Header.Add("Content-Type", "application/octet-stream")

// Send the request to the local web service
resp, err := client.Do(request)
if err != nil {
    return "", err
}

The response contains a Body attribute and that attribute is unmarshalled to a variable of type interface. That one is marshalled with indentation to a byte slice (b) which is returned by the function as a string:

var response interface{}
err = json.Unmarshal(respBody, &response)
if err != nil {
    return "", err
}
b, err := json.MarshalIndent(response, "", "\t")

Now you can use a picture like the one below:

Is he smiling?

Here are some parts of the input, following the command
detectface -image smiling.jpg

Emotion is clearly happiness with additional features such as age, gender, hair color, etc…

[
{
"faceAttributes": {
"accessories": [],
"age": 33,
"blur": {
"blurLevel": "high",
"value": 1
},
"emotion": {
"anger": 0,
"contempt": 0,
"disgust": 0,
"fear": 0,
"happiness": 1,
"neutral": 0,
"sadness": 0,
"surprise": 0
},
"exposure": {
"exposureLevel": "goodExposure",
"value": 0.71
},
"facialHair": {
"beard": 0.6,
"moustache": 0.6,
"sideburns": 0.6
},
"gender": "male",
"glasses": "NoGlasses",
"hair": {
"bald": 0.26,
"hairColor": [
{
"color": "black",
"confidence": 1
}],
"faceId": "b6d924c1-13ef-4d19-8bc9-34b0bb21f0ce",
"faceRectangle": {
"height": 1183,
"left": 944,
"top": 167,
"width": 1183
}
}
]

That’s it! Give the Face API container a go with the tool. You can get it here: https://github.com/gbaeke/msface-go/releases/tag/v0.0.1 (Windows)

Using the Microsoft Face API to detect emotions in photos and video

In a previous post, I blogged about detecting emotions with the ONNX FER+ model. As an alternative, you can use cloud models hosted by major cloud providers such as Microsoft, Amazon and Google. Besides those, there are many other services to choose from.

To detect facial emotions with Azure, there is a Face API in two flavours:

  • Cloud: API calls are sent to a cloud-hosted endpoint in the selected deployment region
  • Container: API calls are sent to a container that you deploy anywhere, including the edge (e.g. IoT Edge device)

To use the container version, you need to request access via this link. In another blog post, I already used the Text Analytics container to detect sentiment in a piece of text.

Note that the container version is not free and needs to be configured with an API key. The API key is obtained by deploying the Face API in the cloud. Doing so generates a primary and secondary key. Be aware that the Face API container, like the Text Analytics container, needs connectivity to the cloud to ensure proper billing. It cannot be used in completely offline scenarios. In short, no matter the flavour you use, you need to deploy the Face API. It will appear in the portal as shown below:

Deployed Face API (part of Cognitive Services)

Using the API is a simple matter. An image can be delivered to the API in two ways:

  • Link: just provide a URL to an image
  • Octet-stream: POST binary data (the image’s bytes) to the API

In the Go example you can find on GitHub, the second approach is used. You simply open the image file (e.g. a jpg or png) and pass the byte array to the endpoint. The endpoint is in the following form for emotion detection:

https://westeurope.api.cognitive.microsoft.com/face/v1.0/detect?returnFaceAttributes=emotion

Instead of emotion, you can ask for other attributes or a combination of attributes: age, gender, headPose, smile, facialHair, glasses, emotion, hair, makeup, occlusion, accessories, blur, exposure and noise. You simply add them together with +’s (e.g. emotion+age+gender). When you add attributes, the cost per call will increase slightly as will the response time. With the additional attributes, the Face API is much more useful than the simple FER+ model. The Face API has several additional features such as storing and comparing faces. Check out the documentation for full details.

To detect emotion in a video, the sample at https://github.com/gbaeke/emotion/blob/master/main.go contains some commented out code in the import section and around line 100 so you can use the Face API via the github.com/gbaeke/emotion/faceapi/msface package’s GetEmotion() function instead of the GetEmotion() function in the code. Because we have the full webcam image and face in an OpenCV mat, some extra code is needed to serialize it to a byte stream in a format the Face API understands:

encodedImage, _ := gocv.IMEncode(gocv.JPEGFileExt, face)       
emotion, err = msface.GetEmotion(bytes.NewReader(encodedImage))

In the above example, the face region detected by OpenCV is encoded to a JPG format as a byte slice. The byte slice is simply converted to an io.Reader and handed to the GetEmotion() function in the msface package.

When you use the Face API to detect emotions in a video stream from a webcam (or a video file), you will be hitting the API quite hard. You will surely need the standard tier of the API which allows you to do 10 transactions per second. To add face and emotion detection to video, the solution discussed in Detecting Emotions in FER+ is a better option.

Detecting emotions with FER+

In an earlier post, I discussed classifying images with the ResNet50v2 model. Azure Machine Learning Service was used to create a container image that used the ONNX ResNet50v2 model and the ONNX Runtime for scoring.

Continuing on that theme, I created a container image that uses the ONNX FER+ model that can detect emotions in an image. The container image also uses the ONNX Runtime for scoring.

You might wonder why you would want to detect emotions this way when there are many services available that can do this for you with a simple API call! You could use Microsoft’s Face API or Amazon’s Rekognition for example. While those services are easy to use and provide additional features, they do come at a cost. If all you need is basic detection of emotions, using this FER+ container is sufficient and cost effective.

Azure Face API (image from Microsoft website)

A notebook to create the image and deploy a container to Azure Container Instances (ACI) can be found here. The notebook uses the Azure Machine Learning SDK to register the model to an Azure Machine Learning workspace, build a container image from that model and deploy the container to ACI. The scoring script score.py is shown below.

score.py

The model expects an 64×64 gray scale image of a face in an array with the following dimensions: [1][1][64][64]. The output is JSON with a results array that contains the probabilities for each emotion and a time field with the inference time.

The emotion probabilities are in this order:

0: "neutral", 1: "happy", 2: "surprise", 3: "sadness", 4: "anger", 5: "disgust", 6: "fear", 7: "contempt

To actually capture the emotions, I wrote a small demo program in Go that uses OpenCV (via GoCV). You can find it on GitHub: https://github.com/gbaeke/emotion. You will need to install OpenCV and GoCV. Find the instructions here: https://gocv.io/getting-started/linux/. There are similar instructions for Mac and Windows but I have not tried those

The program is still a little rough around the edges but it does the trick. The scoring URI is hard coded to http://localhost:5002/score. With Docker installed, use the following command to install the scoring container:

 docker run -d -p 5002:5001 gbaeke/onnxferplus

Have fun with it!

ResNet50v2 classification in Go with a local container

To quickly go to the code, go here. Otherwise, keep reading…

In a previous blog post, I wrote about classifying images with the ResNet50v2 model from the ONNX Model Zoo. In that post, the container ran on a Kubernetes cluster with GPU nodes. The nodes had an NVIDIA v100 GPU. The actual classification was done with a simple Python script with help from Keras and Numpy. Each inference took around 25 milliseconds.

In this post, we will do two things:

  • run the scoring container (CPU) on a local machine that runs Docker
  • perform the scoring (classification) in Go

Installing the scoring container locally

I pushed the scoring container with the ONNX ResNet50v2 image to the following location: https://cloud.docker.com/u/gbaeke/repository/docker/gbaeke/onnxresnet50v2. Run the container with the following command:

docker run -d -p 5001:5001 gbaeke/onnxresnet50

The container will be pulled and started. The scoring URI is on http://localhost:5001/score.

Note that in the previous post, Azure Machine Learning deployed two containers: the scoring container (the one described above) and a front-end container. In that scenario, the front-end container handles the HTTP POST requests (optionally with SSL) and route the request to the actual scoring container.

The scoring container accepts the same payload as the front-end container. That means it can be used on its own, as we are doing now.

Note that you can also use IoT Edge, as explained in an earlier post. That actually shows how easy it is to push AI models to the edge and use them locally, befitting your business case.

Scoring with Go

To actually classify images, I wrote a small Go program to do just that. Although there are some scientific libraries for Go, they are not really needed in this case. That means we do have to create the 4D tensor payload and interpret the softmax result manually. If you check the code, you will see that is not awfully difficult.

The code can be found in the following GitHub repository: https://github.com/gbaeke/resnet-score.

Remember that this model expects the input as a 4D tensor with the following dimensions:

  • dimension 0: batch (we only send one image here)
  • dimension 1: channels (one for each; RGB)
  • dimension 2: height
  • dimension 3: width

The 4D tensor needs to be serialized to JSON in a field called data. We send that data with HTTP POST to the scoring URI at http://localhost:5001/score.

The response from the container will be JSON with two fields: a result field with the 1000 softmax values and a time field with the inference time. We can use the following two structs for marshaling and unmarshaling

Input and output of the model

Note that this model expects pictures to be scaled to 224 by 224 as reflected by the height and width dimensions of the uint8 array. The rest of the code is summarized below:

  • read the image; the path of the image is passed to the code via the -image command line parameter
  • the image is resized with the github.com/disintegration/imaging package (linear method)
  • the 4D tensor is populated by iterating over all pixels of the image, extracting r,g and b and placing them in the BCHW array; note that the r,g and b values are uint16 and scaled to fit in a uint8
  • construct the input which is a struct of type InputData
  • marshal the InputData struct to JSON
  • POST the JSON to the local scoring URI
  • read the HTTP response and unmarshal the response in a struct of type OutputData
  • find the highest probability in the result and note the index where it was found
  • read the 1000 ImageNet categories from imagenet_class_index.json and marshal the JSON into a map of string arrays
  • print the category using the index with the highest probability and the map

What happens when we score the image below?

What is this thing?

Running the code gives the following result:

$ ./class -image images/cassette.jpg

Highest prob is 0.9981583952903748 at 481 (inference time: 0.3309464454650879 )
Probably [n02978881 cassette

The inference time is 1/3 of a second on my older Linux laptop with a dual-core i7.

Try it yourself by running the container and the class program. Download it from here (Linux).

Recognizing images with Azure Machine Learning and the ONNX ResNet50v2 model

Featured image from: https://medium.com/comet-app/review-of-deep-learning-algorithms-for-object-detection-c1f3d437b852

In a previous post, I discussed the creation of a container image that uses the ResNet50v2 model for image classification. If you want to perform tasks such as localization or segmentation, there are other models that serve that purpose. The image was built with GPU support. Adding GPU support was pretty easy:

  • Use the enable_gpu flag in the Azure Machine Learning SDK or check the GPU box in the Azure Portal; the service will build an image that supports NVIDIA cuda
  • Add GPU support in your score.py file and/or conda dependencies file (scoring script uses the ONNX runtime, so we added the onnxruntime-gpu package)

In this post, we will deploy the image to a Kubernetes cluster with GPU nodes. We will use Azure Kubernetes Service (AKS) for this purpose. Check my previous post if you want to use NVIDIA V100 GPUs. In this post, I use hosts with one V100 GPU.

To get started, make sure you have the Kubernetes cluster deployed and that you followed the steps in my previous post to create the GPU container image. Make sure you attached the cluster to the workspace’s compute.

Deploy image to Kubernetes

Click the container image you created from the previous post and deploy it to the Kubernetes cluster you attached to the workspace by clicking + Create Deployment:

Starting the deployment from the image in the workspace

The Create Deployment screen is shown. Select AKS as deployment target and select the Kubernetes cluster you attached. Then press Create.

Azure Machine Learning now deploys the containers to Kubernetes. Note that I said containers in plural. In addition to the scoring container, another frontend container is added as well. You send your requests to the front-end container using HTTP POST. The front-end container talks to the scoring container over TCP port 5001 and passes the result back. The front-end container can be configured with certificates to support SSL.

Check the deployment and wait until it is healthy. We did not specify advanced settings during deployment so the default settings were chosen. Click the deployment to see the settings:

Deployment settings including authentication keys and scoring URI

As you can see, the deployment has authentication enabled. When you send your HTTP POST request to the scoring URI, make sure you pass an authentication header like so: bearer primary-or-secondary-key. The primary and secondary key are in the settings above. You can regenerate those keys at any time.

Checking the deployment

From the Azure Cloud Shell, issue the following commands in order to list the pods deployed to your Kubernetes cluster:

  • az aks list -o table
  • az aks get-credentials -g RESOURCEGROUP -n CLUSTERNAME
  • kubectl get pods
Listing the deployed pods

Azure Machine Learning has deployed three front-ends (default; can be changed via Advanced Settings during deployment) and one scoring container. Let’s check the container with: kubectl get pod onnxgpu-5d6c65789b-rnc56 -o yaml. Replace the container name with yours. In the output, you should find the following:

resources:
limits:
nvidia.com/gpu: "1"
requests:
cpu: 100m
memory: 500m
nvidia.com/gpu: "1"

The above allows the pod to use the GPU on the host. The nvidia drivers on the host are mapped to the pod with a volume:

volumeMounts:
- mountPath: /usr/local/nvidia
name: nvidia

Great! We did not have to bother with doing this ourselves. Let’s now try to recognize an image by sending requests to the front-end pods.

Recognizing images

To recognize an image, we need to POST a JSON payload to the scoring URI. The scoring URI can be found in the deployment properties in the workspace. In my case, the URI is:

http://23.97.218.34/api/v1/service/onnxgpu/score

The JSON payload needs to be in the below format:

{"data": [[[[143.06100463867188, 130.22100830078125, 122.31999969482422, ... ]]]]} 

The data field is a multi-dimensional array, serialized to JSON. The shape of the array is (1,3,224,224). The dimensions correspond to the batch size, channels (RGB), height and width.

You only have to read an image and put the pixel values in the array! Easy right? Well, as usual the answer is: “it depends”! The easiest way to do it, according to me, is with Python and a collection of helper packages. The code is in the following GitHub gist: https://gist.github.com/gbaeke/b25849f3813e9eb984ee691659d1d05a. You need to run the code on a machine with Python 3 installed. Make sure you also install Keras and NumPy (pip3 install keras / pip3 install numpy). The code uses two images, cat.jpg and car.jpg but you can use your own. When I run the code, I get the following result:

Using TensorFlow backend.
channels_last
Loading and preprocessing image… cat.jpg
Array shape (224, 224, 3)
Array shape afer moveaxis: (3, 224, 224)
Array shape after expand_dims (1, 3, 224, 224)
prediction time (as measured by the scoring container) 0.025304794311523438
Probably a: Egyptian_cat 0.9460222125053406
Loading and preprocessing image… car.jpg
Array shape (224, 224, 3)
Array shape afer moveaxis: (3, 224, 224)
Array shape after expand_dims (1, 3, 224, 224)
prediction time (as measured by the scoring container) 0.02526378631591797
Probably a: sports_car 0.948998749256134

It takes about 25 milliseconds to classify an image, or 40 images/second. By increasing the number of GPUs and scoring containers (we only deployed one), we can easily scale out the solution.

With a bit of help from Keras and NumPy, the code does the following:

  • check the image format reported by the keras back-end: it reports channels_last which means that, by default, the RGB channels are the last dimensions of the image array
  • load the image; the resulting array has a (224,224,3) shape
  • our container expects the channels_first format; we use moveaxis to move the last axis to the front; the array now has a (3,224,224) shape
  • our container expects a first dimension with a batch size; we use expand_dims to end up with a (1,3,224,224) shape
  • we convert the 4D array to a list and construct the JSON payload
  • we send the payload to the scoring URI and pass an authorization header
  • we get a JSON response with two fields: result and time; we print the inference time as reported by the container
  • from keras.applications.resnet50, we use the decode_predictions class to process the result field; result contains the 1000 values computed by the softmax function in the container; decode_predictions knows the categories and returns the first five
  • we print the name and probability of the category with the highest probability (item 0)

What happens when you use a scoring container that uses the CPU? In that case, you could run the container in Azure Container Instances (ACI). Using ACI is much less costly! In ACI with the default setting of 0.1 CPU, it will take around 2 seconds to score an image. Ouch! With a full CPU (in ACI), the scoring time goes down to around 180-220ms per image. To achieve better results, simply increase the number of CPUs. On the Standard_NC6s_v3 Kubernetes node with 6 cores, scoring time with CPU hovers around 60ms.

Conclusion

In this post, you have seen how Azure Machine Learning makes it straightforward to deploy GPU scoring images to a Kubernetes cluster with GPU nodes. The service automatically configures the resource requests for the GPU and maps the NVIDIA drivers to the scoring container. The only thing left to do is to start scoring images with the service. We have seen how easy that is with a bit of help from Keras and NumPy. In practice, always start with CPU scoring and scale out that solution to match your requirements. But if you do need GPUs for scoring, Azure Machine Learning makes it pretty easy to do so!