Getting familiar with Semantic Segmentation

[ashenfernando1]

Hey guys, I will provide the labeled eye region data next week, but let's get familiar with the process of semantic segmentation.

I used the following tutorial myself and employed it in the current workflow: https://www.tensorflow.org/tutorials/images/segmentation, but you may be more partial to using/learning PyTorch and that's fine too.

In my experience, the difficult parts come from creating your own input data pipeline, as typically it isn't quite straight forward as examples in tutorials, but working through them would be a good first step and will hopefully give you an appreciation for some of the components of a segmentation workflow.

Of course, the network architecture is a highly customizable object and will be crucial to reducing the size of the model/number of parameters, but more on that throughout the semester.

Those of who you who are interested in the computer vision/segmentation workflow, feel free to take a stab at the tutorial and report on Feb 20th.

[ashenfernando1]

Did you guys find good pytorch resources for learning segmentation? @sankalpshubham @peterthoang

[peterthoang]

Did you guys find good pytorch resources for learning segmentation? @sankalpshubham @peterthoang

Hey Ash, I've been looking at this tutorial: https://albumentations.ai/docs/examples/pytorch_semantic_segmentation/ It uses the same dataset as the tensorflow tutorial, but implemented with pytorch and albumentations.