taesunwhang/MVAN-VisDial

the result of different features

Opened this issue · 3 comments

Thank you for the code. and I have one question. Is the result of using the feature dan_faster_rcnn_x101 and the feature faster_rcnn_x101 very different?

Hello, we experimented by using those two faster rcnn image features, but similar results were obtained (a bit higher score was obtained from faster_rcnn_x101). The results of our paper is the case we use dan_faster_rcnn_x101 (to obtain bounding boxes of the objects in the image).

Hello, we experimented by using those two faster rcnn image features, but similar results were obtained (a bit higher score was obtained from faster_rcnn_x101). The results of our paper is the case we use dan_faster_rcnn_x101 (to obtain bounding boxes of the objects in the image).

thank you

FileNotFoundError: [Errno 2] Unable to open file (unable to open file: name = 'MVAN-VisDial/features_dan_faster_rcnn_x101_train.h5

image
Is the train_btmup_f.hdf5 file the features_dan_faster_rcnn_x101_train.h5?
Does features_dan_faster_rcnn_x101_train.h5 contain bounding boxes information? Where can I download it?

I didn't find it in the link you specified https://github.com/yuleiniu/rva