search

jssprz / visual_syntactic_embedding_video_captioning

Source code of the paper titled *Improving Video Captioning with Temporal Composition of a Visual-Syntactic Embedding*
MIT License
29 stars 8 forks source link

Could you give more details about the semantic concept (SC) detector? #7

Open RyanLiut opened 2 years ago

RyanLiut commented 2 years ago

Hi,

Thank you for your code. Could you give more details about the semantic concept (SC) detector? Like, how to form the SC vocabulary? Since you only release `cnn_sem_globals' which is the probabilities of SL (400d), I wonder how the 400d corresponds to the word in SC vocabulary?

Thank you.

jssprz commented 2 years ago

Sure! the number of concepts to be classified in a multi-label approach is not the most important part. The most relevant aspect is what kind of words will be considered as concepts. To determine if we considered the most used NOUN, ADJ, VERB that was firstly determined by a universal POS tagging method.