search

nickgkan / butd_detr

Code for the ECCV22 paper "Bottom Up Top Down Detection Transformers for Language Grounding in Images and Point Clouds"
Other
74 stars 11 forks source link

Why use predicted spans instead of ground truth spans during the training process? #43

Closed Daniellli closed 10 months ago

Daniellli commented 10 months ago

hi, sorry for bothering you again.

Why use predicted spans instead of ground truth spans during the training process such as this line.

thank you for your time

ayushjain1144 commented 10 months ago

Hi, we wanted to train the model to be robust to imperfections in the spans, hence we use the imperfect predicted spans instead of gt spans.