tgc1997 / RMN

IJCAI2020: Learning to Discretely Compose Reasoning Module Networks for Video Captioning
79 stars 12 forks source link

a problem about region_feature file #23

Open pplee2020 opened 2 years ago

pplee2020 commented 2 years ago

the shape of sfeats of msvd_region_feature.h5 is 1970 x 26 x 36 x 5, what's the meaning of the last dimensions?thank you!

tgc1997 commented 2 years ago

the shape of sfeats of msvd_region_feature.h5 is 1970 x 26 x 36 x 5, what's the meaning of the last dimensions?thank you!

the areas of the regions

pplee2020 commented 2 years ago

May I ask what these five dimensions represent respectively? Does that include confidence?Thank you very much!

tgc1997 commented 2 years ago

sfeats are computed by _boxes2sfeat(boxes, im):, confidences are not included.