search

xuewyang / Fashion_Captioning

ECCV2020 paper: Fashion Captioning: Towards Generating Accurate Descriptions with Semantic Rewards. Code and Data.
Other
81 stars 13 forks source link

Request for Information on Dataset Division and ID Pairing #14

Open houjie8888 opened 1 year ago

houjie8888 commented 1 year ago

Hello, I would like to know how you divided the dataset into training and testing sets, as the .h5 files stored in Google Drive do not appear to have paired IDs and text information, nor any JSON file regarding the image IDs in the training set. Would you be able to provide me with one?

xuewyang commented 1 year ago

Hi, these are the only data that I have right now. It has train, val, test splits. https://drive.google.com/drive/folders/1cgdHt8AlBukmPhuSzUTPszYPXAYmg6gy

houjie8888 commented 1 year ago

Hi, these are the only data that I have right now. It has train, val, test splits. https://drive.google.com/drive/folders/1cgdHt8AlBukmPhuSzUTPszYPXAYmg6gy

Thank you for your response. I'd like to confirm if the order of the .h5 and .json files are aligned. As the .h5 file doesn't contain the image ID attribute, I want to directly retrieve each image along with its corresponding description. How can I accomplish this task?

omgwenxx commented 5 months ago

I don't know if this is still relevant to you, but I used the FICDataset class to load in the files from the Gdrive. As far as I saw their positions align, so each index codes a pairing of image and caption.

However, you will notice, that captions are repeated if the clothing item stays the same and the image just shows different positions. Also, be aware images have different numbers of captions.