Tracking Custom Objects - TensorFlow Object Detection API Tutorial p.3

TL;DR

Learn how to track a custom object using the Tensorflow Object Detection API, including steps for training a model and labeling images.

Transcript

what is going on everybody and welcome to part three of our tensor flow object detection API tutorials series in this video and the subsequent few videos we're going to be covering how to track a custom object whatever object we could possibly want with this API so using the API and the pre trained models and all that it's actually super simple as ... Read More

Key Insights

• ⏮️ The Tensorflow Object Detection API allows for tracking custom objects using pre-trained models.
• 🚂 It is more efficient to fine-tune a pre-trained model than to train a model from scratch.
• ❓ Gathering and labeling images is a crucial part of the tracking process.
• ⌛ The process of training and exporting a model can be time-consuming.
• 🚄 The API offers a trade-off between accuracy and speed when choosing a pre-trained model.
• ❓ Mixing up image sizes and orientations is recommended for better tracking accuracy.
• ❓ The Label Image program can be used for annotating images.

Questions & Answers

Q: What is the purpose of this tutorial series?

The purpose of this tutorial series is to teach users how to track custom objects using the Tensorflow Object Detection API.

Q: Is training a model from scratch recommended?

No, it is not recommended to train a model from scratch. It is better to use a pre-trained model and fine-tune it for your custom object.

Q: How many images of the custom object are needed for tracking?

It is recommended to have between 100 and 500 images of the custom object for tracking. The more images, the better, but it can become tedious to label and handle a large number of images.

Q: What are the steps involved in tracking a custom object?

The steps involved in tracking a custom object include gathering and labeling images, splitting them into train and test samples, generating an ATF record, setting up a configuration file, training a model, and exporting the graph.

Summary & Key Takeaways

• This tutorial series covers how to track custom objects using the Tensorflow Object Detection API.

• The API makes it simple to track objects, but adding a custom object requires more steps.

• The process involves gathering and labeling images, splitting them into train and test samples, generating an ATF record, training a model, and exporting the graph for classification.