This is an implementation of Bridging by Word: Image-Grounded Vocabulary Construction for Visual Captioning based on pytorch 1.0.
-
Install python3.6 and pytorch 1.0.
-
Download the image data of MS COCO to Data/raw/img.
-
Data processing. You need to create a folder in Data/train with the name of coco_v? to store the processed data.
cd Code
python BuildImgOrderClusterVocab.py- Model training.
cd Code
python Trainer.py