As part of updates to arachn.io, I’ve started tinkering with object detection machine learning models. During my experiments on AWS SageMaker, I found that AutoPilot does not support object detection models, so I had train using notebooks. As a result, I hit some “gotchas” fine-tuning TensorFlow Object Detection models. While this notebook works a treat on its own training data (at least when run through SageMaker studio), this discussion will focus on things I learned while trying to run it on my own data on August 31, 2024.
Continue reading “AWS SageMaker Object Detection Training Gotchas”
from Andy Boothe