search

Snowstormfly / Cross-modal-retrieval-MLAGT

Multi-Level and Attention-Guided Tokenization
7 stars 0 forks source link

Request Code to generate the table of the seen and unseen classes #3

Closed LordsMikel closed 5 months ago

LordsMikel commented 5 months ago

Hi,

I admire your repository and research work; it is fantastic. However, I couldn't find the code to generate the tables of seen and unseen classes to test the model. Could you please provide the code to generate them and test the model? I would greatly appreciate it.

Best wishes, Mikel

Snowstormfly commented 5 months ago

Hi, Mikel

In our experiments, we did not use any code to generate "seen" and "unseen" classes. The division of "seen" and "unseen" classes was done manually based on the classification method in Table 3 of the paper, where we divided the 20 categories of the RSketch_Ext dataset into 4-folds for training and testing to evaluate the performance of our model. Therefore, if you want to test the model's effectiveness, you should also manually modify the categories of the training and testing sets to meet the corresponding requirements before testing.

Best wishes,

Bo Yang