KDD Cup 2020 Challenges for Modern E-Commerce Platform: Multimodalities Recall

Team: MTDP_CVA

Members: Kai Zuo, Chao Ma, Dongshuai Li, Zuo Cao, Xing Xu

Introduction

the official have prepared the real-scenario multimodal data from the mobile Taobao. The dataset consists of Taobao search queries and product image features, which is organized into a query-based multimodal retrieval task.The whole training set consists of 3M pairs of query and ground-truth product image features. These sample pairs can be taken as the positive examples to train your retrieval model. For each product image, the official release the features, locations and classification labels of each detected object. See more details: https://tianchi.aliyun.com/competition/entrance/231786/information

Codes

kdd_evaluate_ensemble 
  |--external_resources
  |--user_data
     |--bert_config.json
     |--bert_model.ckpt.data-00000-of-00001
     |--bert_model.ckpt.index
     |--pytorch_model.bin
     |--query_labels.txt
     |--vocab.txt
  |--data
     |--multimodal_labels.txt
     |--testA
        |--testA.tsv
     |--testB
        |--testB.tsv
     |--train
     |--valid
  |--models
     |--BEST.pth
     |--ImageBertKDD.ckpt-85002.data-00000-of-00001
     |--ImageBertKDD.ckpt-85002.index
     |--ImageBertKDD.ckpt-85002.meta
     |--model_attention_kdd_am_word_match_finetune_valid.ckpt-251.data-00000-of-00001
     |--model_attention_kdd_am_word_match_finetune_valid.ckpt-251.index
     |--model_attention_kdd_am_word_match_finetune_valid.ckpt-251.meta
  |--prediction_result
  |--code

cuda 9.0.176 cudnn 7.1.4

下载模型文件 链接:https://pan.baidu.com/s/1tkmpLdF_uguB7VFPNn0cYQ 密码:gwdc

运行脚本 python2 code/main.py