Is the refcoco-trained model available?

jackroos / VL-BERT

Code for ICLR 2020 paper "VL-BERT: Pre-training of Generic Visual-Linguistic Representations".

MIT License

735 stars 110 forks source link

Is the refcoco-trained model available? #67

Closed st2yang closed 3 years ago

st2yang commented 3 years ago

Hello,

It seems that only generic pre-trained models are available, and we have to fine-tune the model on the specific task, right? I am wondering if the refcoco-trained model is available for evaluation?

Thanks, st2yang

jackroos commented 3 years ago

Yes, you need to finetune them on downstream tasks. And I am sorry these models are not available since I didn't store them until now.

MI-WENYU commented 2 years ago

Hello, st2yang

Can you reproduce the results in the author's paper when fine-tuning the program (for the Refcoco dataset)? When I fine-tuned the test, the accuracy was only about 44 percent

Thanks

MI-WENYU commented 2 years ago

Hello,

It seems that only generic pre-trained models are available, and we have to fine-tune the model on the specific task, right? I am wondering if the refcoco-trained model is available for evaluation?

Thanks, st2yang

Hello, st2yang

Can you reproduce the results in the author's paper when fine-tuning the program (for the Refcoco dataset)? When I fine-tuned the test, the accuracy was only about 44 percent

Thank

st2yang commented 2 years ago

@MI-WENYU Yes

MI-WENYU commented 2 years ago

@MI-WENYU Yes @st2yang Does the reproduction process use the original parameters? I have been unable to achieve the results in the paper. Could you please share your fine-tuned model with me? I would be very grateful.