search

OFA-Sys / OFA

Official repository of OFA (ICML 2022). Paper: OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework
Apache License 2.0
2.4k stars 247 forks source link

Why are all the sentences generated by image captioning the same sentence? #294

Open Swieo opened 1 year ago

Swieo commented 1 year ago

My dataset is medical images (x-rays), but after fine-tuning ， the generated captions of all the images are the same sentence.Is there something else that needs to be modified?Hope the author can give me some advice.

JustinLin610 commented 1 year ago

Any more details? You can first check whether you can reproduce the performance following our instructions for finetuning on coco captions. If everything is alright, you can check your code or provide a colab notebook for us to follow up.

Swieo commented 1 year ago

Any more details? You can first check whether you can reproduce the performance following our instructions for finetuning on coco captions. If everything is alright, you can check your code or provide a colab notebook for us to follow up.

Because I just have one not 5 captions about an image, so I didn't do the stage 2 training, is it because of that?

levietthinh commented 11 months ago

My dataset is medical images (x-rays), but after fine-tuning ， the generated captions of all the images are the same sentence.Is there something else that needs to be modified?Hope the author can give me some advice.

Hi, did you find out the reasons and solutions?