giangnguyen2412 / contcap Goto Github PK

View Code? Open in Web Editor NEW

4.0 4.0 3.0 1.39 MB

Code for: Contcap - A scalable framework for continual image captioning

Python 97.86% Shell 2.14%

contcap's People

Contributors

Stargazers

Watchers

Forkers

prishruit andrei-inacio jayashree-pougajendy

contcap's Issues

Google drive links don't work.

The google drive links for the coco captions and the models doesn't work, can you please update them.

Inference on single wild video using pre-trained models

Hey @giangnguyen2412
Thanks for open sourcing such awesome work!

I was wondering how would someone go about using some of your pre-trained models for inference on single wild videos?

Please share the vocabulary file too.

Can you please also share the vocabulary file that is mentioned under section Training and inference of 2to21 under step 2, also when I trained from scratch my results for this process i.e after training the base task (2to21) and inferring on the same task are as follows :
CIDEr : 0.431
SPICE : 0.095
ROUGE_L : 0.428
METEOR : 0.176
Bleu_1 : 0.564
Can you confirm that you got similar results and share results after training the base task of 2to21.

Originally posted by @Prishruit in #2 (comment)

Recommend Projects