Please open an issue if you have questions or problems with the dataset.
The 2017 competition, sponsored by Google, is part of the FGVC^4 workshop at CVPR.
We are using Kaggle to host the leaderboard. Checkout the competition page here.
Data Released | April 5, 2017 |
Submission Server Open | June 1 2017 |
Submission Deadline | July 7, 2017 |
Winners Announced | July 21, 2017 |
iMaterialist is a fine-grained attribute recognition challenge for product images. With the growing popularity of online shopping, being able to recognize products from pictures is an important but challenging problem. The recognition is challenging because, for the same product, a picture can be taken under different lighting and viewing angle, various backgrounds, and different level of occlusion. One central piece for recognizing a product is to understand and obtain the attributes information of a product. The attributes describes properties of a product. For example, in the figure below, the pair of shoes have the attributes described on the right part of the image.
In this competition, you are challenged to develop recognition / classification algorithms which accurately assign attribute labels for apparel items in the image. By participating in this competition, you will contribute to the state-of-the-art computer vision and machine research and setting benchmarks for fine-grained attribute product recognition.
For more information on the Evaluation Metric, Data / Submission Format, Rules and participating the iMaterialist Competition, please see the Kaggle challenge page.