SIGIR2020Challenge

KO | EN

gitlite — search

#python #react #ai #llm #golang #docker #macos #java #windows #awesome-list #security #mcp #ios #openai

SIGIR2020Challenge

(★ 21)

The 1st place solution for SIGIR 2020 E-Commerce Workshop Multimodal Product Classification Challenge

SIGIR2020Challenge 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# SIGIR 2020 E-Commerce Workshop Data Challenge (Multi-modal Classification Task) 
## Task Intro

[This challenge](https://sigir-ecom.github.io/data-task.html) focuses on the topic of large-scale multi-modal (text and image) classification, where the goal is to predict each product’s type code as defined in the catalog of Rakuten France. To be specific, given a training set of products and their product type codes, predict the corresponding product type codes for an unseen held out test set of products. The systems are free to use the available textual titles and/or descriptions whenever available and additionally the images to allow for true multi-modal learning.

## Dataset 
The organizer released approximately 99K product listings in tsv format, including 84,916 samples for training, 937 samples for phase 1 testing and 8435 samples for phase 2 testing. The training dataset consists of product titles, descriptions, images and their corresponding product type codes. There are 27 product categories in the training dataset and the number of product samples in each category ranges from 764 to 10,209. The frequency distribution of categories in the training dataset is shown in the following figure.
![category frequency distribution](https://raw.githubusercontent.com/Wang-Shuo/SIGIR2020Challenge/master/assets/data_dist.png)

## Overview
We employed a decision-level fusion scheme to leverage mulitmodal product information. Specifically, the modal-specific classifiers are built from textual and image modal data respectively in the first stage. Then the late fusion strategy is learned from the class probabilities predicted by each modal classifier in the second stage. The overview of the proposed method is shown in the following figure.
![model overview](https://raw.githubusercontent.com/Wang-Shuo/SIGIR2020Challenge/master/assets/overview.png)

## Reference

SIGIR2020Challenge

SIGIR2020Challenge 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

SIGIR2020Challenge

SIGIR2020Challenge 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README