A Mask R-CNN Keras implementation with Modanet annotations on the Paperdoll dataset
This release fixes a problem mentioned here: https://github.com/eBay/modanet/issues/8#issuecomment-526855697
Now, as you can see on the downloads of this release, it is included in this repo. Along with other small fixes to the bash script to download the necessary stuff. It no longer requires LFS (Git Large File Storage) if you fast_download it!
Just a quick fix for instagram implementation that crashed when opening an empty label.
This release fixes a few things and implements a new image saving method, better than matplotlib and with greatly reduced images file size.
Just some fixes, stable-ish release now
EDIT: now live on https://pypi.org/project/maskrcnn-modanet/#description
Now you can download it in just a few mins!
It includes an instagram analyzer tool that scans through all the images of an instagram user and shows some statistics about what type of apparel the user is wearing and even displays the instances of them, if you request it.
Say you want to quickly find what skirt (or footwear) your instagram star always wears. With this tool you can! And you can also see how often the instagram user shows himself alone in their images, and what he/she usually shares of him (always pictures with shoes? always only the top part?)
It's really nice overall and it's just a simple application of what this technology can do.
If you have a previous version, run the download command again (you can run it in the same folder)
The resnet50_modanet.h5 file attached is after 30 epochs of default settings, with fixed dataset.
Clone the repository to the last master, don't download the source code from here.
Hi all! The next release will give you the fast download method, which shrinks the download from 40GB to just 2-3 GB (at the expense of not having downloaded the whole PaperDoll dataset).