search

Cadene / vqa.pytorch

Visual Question Answering in Pytorch
717 stars 178 forks source link

Performance of MLB #7

Closed gaopeng-eugene closed 7 years ago

gaopeng-eugene commented 7 years ago

Why your MLB baseline is so low?

Cadene commented 7 years ago

You are talking about the experimental results of table1, I guess. https://arxiv.org/pdf/1705.06676.pdf

To point out the performance variation due to the fusion modules, we first compare MUTAN to state-of-the-art bilinear models, under the same experimental framework. We do not use attention models here.

So in those experiments image features are of size 2048 instead of 14x14x2048.

gaopeng-eugene commented 7 years ago

What is the performance of MLB+attn in your code?

Cadene commented 7 years ago

You can download the zip file.