Open hungnt14 opened 2 years ago
One way is to follow the colab, which uses a stronger bottom up feature(ectracted with pytorch). Or you can checkout my Google conceptual repo which provides a docker that includes caffeine bottom up and corresponding captioning code.
Do you support generating captions using model that trained with Bottom up features?
If not, can you give me a hint to predict caption from our raw images using that BU Model?
Thanks a lot for your awesome work!