nqanh / video2command Goto Github PK

Hello,

Again thank you for providing the ResNet50 feature files last time. However I'd actually be more helpful if you can provide the missing script that you used to generate all those feature files. That is, the script that you used to generate all:

• resnet
• vgg19
• resnet50_keras_feature_no_sub_mean
• vgg16_keras_feature_no_sub_mean
  and so on...

One of the keys for training you mentioned during your paper is the way you augment the frames with synthetic imagenet frames. I think this part of the code is also included in your missing script. (This is also the part that interests me the most)

Thank you again in advance!

Hello, Prof. Nguyen.
I want to use the TCN Module of V2C net. It would be very beneficial to me if you can provide the code of the same.

Hello, first of all, this is a very interesting work combing captioning into the robotic application.

I am wondering if any pre-trained models are available for inference? Or would you mind providing the missing pre-extracted features with ResNet50 so that it will be easier to train a model on my own?

Thanks in advance!

Edit: From my review of the code, it seems that the part of the code for extracting image features and packing up (images, captions) into a single pkl file is missing.

Data_io.py file reported an error. Is there any missing file?
IndexError：list index out of range

Hello, Prof. Nguyen. I want to follow this work. My access was denied on Google drive.
How do I get the IIT-V2C dataset? Thanks.

Hello,
Thank you for your answer to the last question. The model has been successfully run.Now I am confused about the video frame extraction method and supplementary frame method. I hope to get your help. Thank you again.