This is a python pipeline, combining different models to analyze clothes of a picture. If the (python) tensorflow code and the overall pipeline are given here, the models haven't been uploaded.
- Install tensorflow (>=1.8, python 2.7 or 3). Tensorflow-gpu recommended.
- Use your own models.
- Update the global variables in main.py (especially for folder and model paths)
- python main.py
This pipeline detects clothes, but you can reuse it for whatever purpose
- We load an image
- We remove the background with deeplab (we extract the persons)
- We detect the objects we want (here, the clothes. We also make sure that we can't detect that a person wears a dress, a top and a t-shirt at the same time!)
- We extract the last layer of an inception model. This is used to analyze similarity between clothes
- We could apply additional classifiers here but as they are not ready yet, we just save everything.