Skip to content

cpheinrich/deeptail

Repository files navigation

deeptail

https://www.kaggle.com/c/whale-categorization-playground

Getting Started

Image Segmentation

  • An optional preprocessing step to improve classification accuracy is to first segment the whale tail from the background (typically sky and water).
  • segmentation/unsupervised\ segmentation.ipynb has a pipeline for segmenting images using the open CV watershed algorithm. Unfortunately, this doesn't always deliver sastisfactory results, so we also include a few supervised segmentation approaches, and some sample training data.
  • Unzip the segmentation/ directories for validation and training, and then you can run commands like in the sample_commands.txt file
  • More info in segmentation/README.md

Resources

ipynbs from Chollet's deep learning with python:

Remarks: Sections 5.2 and 5.3 provide good examples of image classifiers. The main difference in the whale data is that it is not a binary classifier, and as such, we will likely want to one-hot encode the whale_ids into binary vectors. This is discussed in sections 3.6 and 6.1

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •