amirassov/kaggle-imaterialist
A first-place Kaggle solution using Hybrid Task Cascade deep learning model for instance segmentation of fashion product images.
Not currently ranked — collecting fresh signals.
star history
This repository contains a winning solution for the iMaterialist Fashion 2019 competition, implementing a Hybrid Task Cascade architecture with ResNeXt-101-FPN backbone for instance segmentation. It uses PyTorch and mmdetection, trains on COCO pretrained weights, and employs test-time augmentation with multi-scale inference and ensembling to achieve state-of-the-art segmentation performance.
Frequently asked
- What is amirassov/kaggle-imaterialist?
- A first-place Kaggle solution using Hybrid Task Cascade deep learning model for instance segmentation of fashion product images.
- Is kaggle-imaterialist open source?
- Yes — amirassov/kaggle-imaterialist is open source, released under the MIT license.
- What language is kaggle-imaterialist written in?
- amirassov/kaggle-imaterialist is primarily written in Python.
- How popular is kaggle-imaterialist?
- amirassov/kaggle-imaterialist has 484 stars on GitHub.
- Where can I find kaggle-imaterialist?
- amirassov/kaggle-imaterialist is on GitHub at https://github.com/amirassov/kaggle-imaterialist.