Issue with custom object detection in the browser using tfjs

General Discussion
#1

Hello, I was wondering if there is a blog or video tutorial that showed us how to do custom object detection in the browser using the most latest versions of tfjs. I tried kangaroo detector and SKU one, but when I tried to load custom models, it did not work, so I believe there may be a version mismatch issue. Alternatively, is there any way that I can use the old object detection api?

#3

@Mayukhdeep Welcome to Tensorflow Forum!

The kangaroo detector and SKU detector might have version mismatch issues due to being older examples. Here are some alternatives for performing custom object detection in the browser using the latest versions of tfjs:

1. TensorFlow.js Object Detection API:

The latest update to tfjs introduced a native Object Detection API, making it easier to perform custom object detection without relying on external libraries. You can find comprehensive documentation and tutorials here:

2. Model Maker and Task API:

This approach utilizes TensorFlow Lite Model Maker to train a custom model on your device and then deploy it using the Task API for real-time inference in the browser. It works with both image and webcam input.

3. External Libraries:

If you prefer using external libraries, consider checking out:

Using the Old Object Detection API:

While officially deprecated, you can still use the old Object Detection API if absolutely necessary. However, remember that it’s no longer actively maintained and might have compatibility issues with the latest tfjs versions.

Troubleshooting Version Mismatch:

If you encounter version mismatch issues with specific libraries, double-check the exact versions used in the examples and ensure your tfjs and any dependencies are up to date. You can also consider searching for updated forks or alternatives that support the latest tfjs versions.

Using the native Object Detection API or Model Maker with Task API are recommended approaches for custom object detection in the browser with the latest tfjs versions. Remember to choose the method that best suits your skill level and project requirements.

I hope this helps! Let me know if you have any further questions.