CV/ML Toolkit for Self Guided Object Detection with Tensorflow, Sagemaker, and Sagemaker Groundtruth.

This material strives to provide a comprehensive set of notebooks to showcase a standard ML Object detection workflow. The workflow is being leveraged as part of a larger foundational CV/ML framework. This repo was forked from Angela Wang's original work but substantially modified to account for (among other things) a custom Tensorflow image, model optimization through Hyperparameter tuning jobs, and model inference. The conclusion of the excercise produces model artifacts that can be consumed by your application.

Object detection is the process of identifying and localizing objects in an image. A typical object detection solution takes in an image as input and provides a bounding box on the image where an object of interest is, along with identifying what object the box encapsulates.

CV/ML Toolkit Objectives:

• Data collection/procurement by way of capturing static frames from video.
• Curating a good data-set to prepare for labeling
• Dataset labeling techniques with Amazon Sagemaker Groundtruth.
• Launching a custom tensorflow container in your account
• Training against that container with a base model of your choice.
• Running Model tuning jobs to optimize your model across a given range of optimizations.
• Testing the accuracy of your model on new images.
• Export your model for use in your application.

For best results, to get started:

1. Log into your AWS Console, select Sagemaker.
2. Create a new notebook instance (or launch an existing one).
3. Once the notebook instance has loaded, select "Open Jupyter" to get into the Jupyter console.
4. From the console, select New (upper right) and then Terminal.
5. Once the terminal window loads:

sh-4.2$ cd SageMaker
sh-4.2$ git clone git@github.com:RaleighSF/amazon-sagemaker-aws-greengrass-custom-object-detection-model.git

6. Now go back to the Jupyter console (or click on the Jupyter logo upper left).
7. Click into the amazon-sagemaker-aws-greengrass-custom-object-detection-model directory.
8. Finally, launch the CVML_Toolkit.ipynb

Architecture

Sections

License Summary

This sample code is made available under the MIT-0 license. See the LICENSE file.