The goal of this notebook is to show how to develop custom Machine Learning models on SageMaker for your business usecase. You will use the SageMaker built-in object detection algorithm and Tensorflow docker images to detect license plate of a car and recognize the characters of the plate using CNN and computer vision technology.
If you want to refer to the introduction of the blog form, please click this link
This HoL consists of four Labs, each with the following details:
- Generating images and annotation data for ML training
- Image synthesis
- Example structure of annotation file
- Using SageMaker built-in algorithm
- Developing custom Object Detection
- Leveraging Transfer Learning (Resnet-50)
- Composing custom CNN(Convolution Neural Net) architecture with Tensorflow and Keras
- Developing and testing custom Tensorflow script and before running training jobs on SageMaker
- Developing with Tensorflow script mode of SageMaker
- Leveraging distributed training in the Cloud
- 1 click deployment and endpoint hosting
- Elastic Inference
- Invoking endpoint with new data
You may expand the result of this Lab to other usecases such as serial number detection of your product, etc.
Seongmoon Kang