Xiongkun Linghu comments

Results 8 comments of


                                            Xiongkun Linghu

All the results are lower than paper with the default settings and feature pyramid 5,2,1

I mean the results on the arxiv paper. Are the results on the table 7 trained with the auxiliary loss?

Can't achieve the accuracy in the papar

Thanks a lot! I will try it again.

Can't achieve the accuracy in the papar

Ok, I will check the paper for more details about this parameter.

Explanation for data format and issues about data generation

Thanks for your reply. Now I understand the format clearly.

[Feature] Other baseline models

Thanks for the response. I am interested in LEO's baseline, especially the data preprocess and data loader for such a baseline.

[Feature] Other baseline models

Willing to see the released baselines soon.

Encoding referent tokens

Thanks for your quik reply. Another question is that if there are multiple referent tokens in the prompt, how can you distinguish different referent scene queries? In above example, "Describe...

Encoding referent tokens

Thanks for the reply, I finally understand the mechanism.