peternara/dresscode-fashion

Fashion Object Retrieval

This repository contains code that performs a catalogue matching task on a custom apparel dataset with Deep Convolutional Neural Networks (DCNNs) techniques. Please consider looking at the report for a full description of this work.

Catalogue Matching

Given the catalogue of a retailer as a dataset of photos, the task is to indentificate which of these objects people are wearing in real life photos that we call queries.

The Dataset

The chosen dataset simplfies the task. It consists in all the dresses of the H&M french online catalogue.
The real world photos consists in the models that wear the items. You get associations like these:

Catalogue Item

Corresponding Model (Query)

Getting the dataset

To get get the dataset you should execute get_dataset.py with python3.
It will download the catalogue items in db/robes/cat/ and the corresponding models in db/robes/mod/.
At the time the experiment was made it consisted in 210 items, the photos are of good quality the whole takes about 200Mo.
Important: if you want to re-conduct the following results you should check that the file to_ignore.txt is still adequate to the DB. This variable stores ill-formed examples (where no models wear the item or ill formated jpg files).

Structure of this repo

Here's a short description of each file in this repo:

• report.pdf: main file of this repo, it explains all of what this is about.
• Results_DCNNs.ipynb: main results of our work, using deep neural features in order to retrieve intraclass apparels.
• Results_ORB_BagOfWords.ipynb: implementation of "old-school" Bag-Of-Words method on the same problem for comparison purpose.
• Visualization_SOM.py: various features visualization with the Self-Organizing Map.
• cache.py: caching tools, directly issued from this awesome github repo.
• data_manager.py: routines for data management.
• download.py: models downloading utility, directly issued from this awesome github repo.
• get_dataset.py: our script for dataset gathering.
• inception.py: manipulation of the inception v3 model, directly issued from this awesome github repo.
• rect_tools.py: routines for manipulating rectangles.
• som.py: our implementation of the SOM.
• to_ignore.txt: specify db items to ignore.
• models/: where to store tensorflow models.
• dumps/: where to store dumps of heavy calculations.

License (MIT)

Copyright (c) 2017 by Tristan Cosmo Stérin

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.