Martin Krasser

They were trained as described in https://github.com/krasserm/fairseq-image-captioning/blob/master/README.md#training

The object-based model is significantly better but when I trained the grid-based model long time ago I didn't really tune hyper-parameters. So it may be worth re-training it with hyper-parameters...

Thanks for your efforts on this @adrelino, getting it integrated into this project would be a great addition! I don't think it's critical to have exactly the same weights as...

Sounds great! :+1: on using the detectron2 model with the R101-FPN backbone but I'm not sure if pre-training on MS-COCO train2017 will work in combination with this project. Here, [Karpathy...

Yes, exactly! Thanks for the pointer @ruotianluo, and your hint regarding the importance of training with the attribute head.

@Kyubyong not sure about @adrelino's plans/progress here. I'm still interested in this feature but have other priorities at the moment. I think what @ruotianluo [describes](https://github.com/krasserm/fairseq-image-captioning/issues/9#issuecomment-631227366) is the way to go...

Initial work in [wip-train-inception](https://github.com/krasserm/fairseq-image-captioning/tree/wip-train-inception) branch (still based on fairseq 0.8.0).

Interesting idea but how do you ensure same lock semantics across DB implementations? I did a lot of research and debugging to get instance-level locking right and performance in Derby....

Sure, that's the way to go but what if the underlying database doesn't support the locking protocol used by the FM manager such as lock upgrades, for example? Or what...

Why do you want to support offline Maven2 builds? Was this a request from a client/user? When adding IPF dependencies to the classpath, you'll anyway flat it again (i.e. outside...