Can I count custom object in single image with Amazon Rekognition Custom Labels?

Check this blog that explains how a store adjusts its workforce based on the number of people detected using Amazon Rekognition. Around the middle of the blog there is a lambda function that uses amazon Rekognition to "count number of people in the picture".

Hi,

Thanks for your interest in Rekognition Custom Labels! This is one of the common use cases for Custom Labels. If you have a set of images that contains a specific type of balls you can annotate them using the Custom Labels console. After that you can submit them for training and will get a model tailored to the specific use case. With the model, you can further run real time inference to get the ball detected and thus counting them.

Check out the documentation https://docs.aws.amazon.com/rekognition/latest/customlabels-dg/understanding-custom-labels.html#tm-object-localization for more information.

Thank you!

Hey Nugroho! I ran into a similar problem some time ago, and found it challenging to solve with Rekognition. It's why we're building DirectAI (https://directai.io). We let you create custom labels to annotate images, with zero custom training or deployment. Just tell us what you want to find, and our API will give you annotations out of the box.

I've attached a demo of what this might look like below. Ask to find all paper lanterns, and count how many results are returned in the JSON. You can also tweak the detection and NMS thresholds to tune the model for your specific use case.

Feel free to schedule some time to chat if any of this sounds relevant. We'd love to hear how we can meet your needs.