TagMyOutfit
Tag-My-Outfit is a classifier that predicts the category and the attributes of a piece of clothing viewed in a given image. It is containerized and gRPC enabled so that it can be used standalone or in a pipeline using AI4EU experiments platform.
The prediction model is the Visual Semantic Attention Model (VSAM), and is supervised by automatic pose extraction creating a discriminative feature space. This particular classifier was trained with the open source DeepFashion dataset.
The asset is integrated in a pipeline that can be deployed with one single command. Full description, source code and documentation in the Github repository and the external url.
Github https://github.com/DuarteMRAlves/tag_my_outfit_pipeline
This asset assigns fashion labels to images of clothing as illustrated in the figure below.
It can be used standalone but to increase performance an image pre-processing is required. In particular we provide a series of other assets that together provide a system deployable with a single command using docker-compose or built and orchestrated in the aiexp platform. Figure below illustrated the main modules.
The pre-processing is done by Openpose (available here ) that detects the body to enable a precise crop of the image and consequently improving the precision of the classifier.