search

krasserm / fairseq-image-captioning

Transformer-based image captioning extension for pytorch/fairseq
Apache License 2.0
312 stars 55 forks source link

Support features extracted from detected objects #3

Closed krasserm closed 4 years ago

krasserm commented 4 years ago

This also supports a variable number of object features. All details are in the updated README. This PR also comes with a major refactoring.

krasserm commented 4 years ago

Thanks for the helpful feedback @cstub. The last commit implements all your suggestions. Regarding --no-projection see this comment.