Does Azure Ml support training deep learning models like yolov3, faster R-CNN, Deeplabv3+, Mask R-CNN

Question

Does the azure ml supports training and inference task for deep learning models for object detection, semantic segmentation models.

Answer 1

Hi,

Azure Machine Learning appears to support Yolov3 and Faster R-CNN:
https://azure.microsoft.com/en-us/blog/new-azure-machine-learning-updates-simplify-and-accelerate-the-ml-lifecycle

https://devblogs.microsoft.com/cse/2017/10/24/bird-detection-with-azure-ml-workbench

Best regards,
Leon

Answer 2

Thanks for reaching out. Yes, Azure ML supports deep learning models for object detection using ONNX. Please review this document for more details. Models from many frameworks including TensorFlow, PyTorch, SciKit-Learn, Keras, Chainer, MXNet, MATLAB, and SparkML can be exported or converted to the standard ONNX format. Once the models are in the ONNX format, they can be run on a variety of platforms/devices including Azure Machine Learning services.

Many models including object detection can be represented as ONNX models and can be obtained through the following ways:

1. Train a new ONNX model in Azure Machine Learning or by using automated Machine Learning capabilities
2. Convert existing model from another format to ONNX
3. Get a pre-trained ONNX model from the ONNX Model Zoo
4. Generate a customized ONNX model from Azure Custom Vision service

Furthermore, you can deploy, manage, and monitor your ONNX models in Azure ML. Please check out the following examples on how-to-use-azureml/deployment/onnx in Python as well as samples in other programming languages. Let us know if you have any further questions or concerns. Thanks.