DEV Community

Cover image for SHIFT15M: Multiobjective Large-scale Fashion Dataset with Distributional Shifts
Masanari KIMURA
Masanari KIMURA

Posted on

SHIFT15M: Multiobjective Large-scale Fashion Dataset with Distributional Shifts

Overview of the SHIFT15M dataset.

Hi, everyone.

We present a novel dataset aimed at properly evaluating machine learning models under distributional shifts.

Our SHIFT15M dataset has several good properties:

  • Multiobjective. Each instance in the dataset has several numerical values that can be used as target variables.
  • Large-scale. The SHIFT15M dataset consists of 15million fashion images.
  • Coverage of types of dataset shifts. SHIFT15M contains multiple dataset shift problem settings (e.g., covariate shift or target shift). SHIFT15M also enables the performance evaluation of the model under various magnitudes of dataset shifts by switching the magnitude.

In addition, we provide software to handle SHIFT15M in a very simple way.

If you are interested feel free to check out:

Arxiv: https://arxiv.org/abs/2108.12992

GitHub: https://github.com/st-tech/zozo-shift15m

Top comments (0)