search

fukun07 / neural-image-captioning

Using scene-specific contexts and region-based attention in neural image captioning
MIT License
44 stars 18 forks source link

how are features_1res extracted from resnet? #11

Closed feiliu95 closed 5 years ago

feiliu95 commented 6 years ago

Dear author, I have some confusion. First, how are features_1res extracted from resnet? apply global pooling on the 'res5c' convolutional feature maps? Second, which ResNet network is used for the extraction of visual features? ResNet-101 or the others? Thank you!

fukun07 commented 6 years ago

Hi, we used ResNet-152, and treat the output of layer "pool5" as the feature.