Google CEO Sundar Pichai wrote, “... designing neuralnets is extremely time intensive, and requires an expertise that limits its use to a smaller community of scientists andengineers." Shortly after this Google launched its service AutoML in early 2018.
AutoML aims to enable developers with limited machine learning expertise to train high-quality models specific to theirbusiness needs. The goal of AutoML is to automate all the major repetitive tasks such asfeature selection orhyperparameter tuning. This allows creating more models inless time with improved quality and accuracy.
A basic two step approach to machine learning: First, the model is created by fitting it to the data. Second, the modelis used to predict the output for new (previously unseen) data.
This blog post demonstrates how to get started quickly with AutoML. It will give you a step-by-step tutorial on how tobuilt an Object Detection Model using AutoGluon, with top-notch accuracy. I created aGoogle Colab Notebook with a full example.
In January 2020 Amazon Web Services Inc. (AWS) secretly launched an open-source library calledAutoGluon the library behind Sagemaker Autopilot.
AutoGluon enables developers to write machine learning-based applications that use image, text or tabular data sets withjust a few lines ohttps://www.philschmid.de/static/blog/getting-started-with-automl-and-aws-autogluon/autogluon-sagemaker.pngn/autogluon-sagemaker.png" alt="sagemaker-and-autogluon">
With those tools, AWS has entered the field of managed AutoML Services or MLaas and to competeGoogle with its AutoML service.
What is AutoGluon?
"AutoGluon enables easy-to-use and easy-to-extend AutoML with a focus on deeplearning and real-world applications spanning image, text, or tabular data. Intended for both ML beginners and experts,AutoGluon enables you to... "
quickly prototype deep learning solutionsautomatic hyperparameter tuning, model selection / architecture searchimprove existing bespoke models and data pipelines **
AutoGluon enables you to build machine learning models with only 3 Lines of Code.
from autogluon import TabularPrediction as taskpredictor = task.fit(train_data=task.Dataset(file_path="TRAIN_DATA.csv"), label="PREDICT_COLUMN")predictions = predictor.predict(task.Dataset(file_path="TEST_DATA.csv"))
Currently, AutoGluon can create models for image classification, object detection, text classification, and supervisedlearning with tabular datasets.
We are going to build an Object Detection Model, to detect fruits (apple, orange and banana) on images. I built a smalldataset with around 300 images to achieve a quick training process.You can find the dataset here.
I am using Google Colab with a GPU runtime for this tutorial. If you are not sure how to use a GPU Runtime take a lookhere.
Okay, now let's get started with the tutorial.
Installing AutoGluon
AutoGluon offers different installation packages for different hardware preferences. For more installation instructionstake a look at the AutoGluon Installation Guide here.
The first step is to install AutoGluon with pip and CUDA support.
# Here we assume CUDA 10.0 is installed. You should change the number# according to your own CUDA version (e.g. mxnet-cu101 for CUDA 10.1).!pip install --upgrade mxnet-cu100!pip install autogluon
For AutoGluon to work in Google Colab, we also have to install ipykernel and restart the runtime.
!pip install -U ipykernel
After a successful restart of your runtime you can import autogluon and print out the version.
import autogluon as agfrom autogluon import ObjectDetection as taskprint(ag.__version__)# >>>> '0.0.6'
Loading data and creating datasets
The next step is to load the dataset we use for the object detection task. In the ObjectDetection task from AutoGluon,you can either use PASCAL VOC format or the COCO format by adjusting the format parameter of Dataset() to eithercoco or voc. ThePascal VOC Dataset containstwo directories: Annotations and JPEGImages. TheCOCO dataset isformatted in JSON and is a collection of "info", "licenses", "images", "annotations", "categories".
For training, we are going to use thetiny_fruit_object_detection dataset, which Ibuild. The Dataset contains around 300 images of bananas, apples, oranges or a combination of them together.
We are using 240 images for training, 30 for testing https://www.philschmid.de/static/blog/getting-started-with-automl-and-aws-autogluon/sample-images.pnged-with-automl-and-aws-autogluon/sample-images.png" alt="sample-images">
Using the commands below, we can download and unzip this dataset, which is only 29MB. After this we create ourDataset for train and test with task.Dataset().
The third step is to train our model with the created dataset. In AutoGluon you define your classifier as variable,here detector and define parameters in the fit() function during train-time. For example, you can define atime_limit which automatically stops the training after a certain time. You can define a range for your ownlearning_rate or set the number of epochs. One of the most important parameters is num_trials. This parameterdefines the maximum number of hyperparameter configurations to try out. You can find the full documentation of theconfigurable parameters here.
We are going to train our model for 20 epochs and train 3 different models by setting num_trials=3.
from autogluon import ObjectDetection as taskepochs = 20detector = task.fit(dataset_train, num_trials=3, epochs=epochs, lr=ag.Categorical(5e-4, 1e-4, 3e-4), ngpus_per_trial=1)
As a result, we are getting a chart with the mean average precision (mAP) and the number of epochs. The mAP is a commonmetric to calculate the accuracy of an object detection model.
Our best model (bluehttps://www.philschmid.de/static/blog/getting-started-with-automl-and-aws-autogluon/result-chart.png/static/blog/getting-started-with-automl-and-aws-autogluon/result-chart.png" alt="result-chart">
Evaluating the Model
After finishing training, we are now going to evaluate/test the performance of our model on our test dataset
test_map = detector.evaluate(dataset_test)print(f"The mAP on the test dataset is {test_map[1][1]}")
The mAP on the test dataset is 0.8724113724113725 which is pretty good considering we only training with 240 Imagesand 20 epochs.
Predict an Image
To use our trained model for predicting you can simply run detector.predict(image_path), which will return a tuple(ind) containing the class-IDs of detected objects, the confidence-scores (prob), and the corresponding predictedbounding box locations (loc).
As of the time writing this article, saving an object detection model is not yet implemented in version 0.0.6, butwill be in the next deployed version.
To save your model, you only have to run detector.save()
savefile = 'model.pkl'detector.save(savefile)
Load Model
As of the time writing this article, loading an object detection model is not yet implemented in version 0.0.6, butwill be in the next deployed version.
