Author: Timofey Vilkov
Supervisor: Panchuk Georgii
Formulation of the problem: Creating a text2image platform based on multimodal models for searching similar sneakers by user text description.
Roadmap:
- Searching sites with large number of items (01.01.2023 — 01.03.2023)
- analyze the quality of content
- estimate complexity of parsing
- Implement parser for scraping image and text data from sneaker platforms (05.03.2023 — 15.03.2023)
- write parser
- parse and scrape data
- extract data from raw responses
- collecting images
- EDA (15.03.2023 — 20.03.2023)
- Data subsample labeling (20.03.2023 — 27.03.2023)
- Research multimodal models, сhoosing the best one (27.03.2023 — 20.04.2023)
- define metrics for decision making
- evaluation of selected models using labelled subset of the data
- choosing the best model
- ML system design (27.03.2023 — 15.04.2023)
- ML system implementation (15.04.2023 — 15.05.2023)
- writing code for inference
- hosted app on the server
- create bot
- create base flow for interact with user
- logging base metrics about load on the model
- Inference optimization (15.05.2023 — 31.05.2023)
Data description: Data: parsed shoe items from https://www.ssense.com/.
Used endpoints:
Raw data: raw_data_urls.json
CSV with urls: url_dataset.csv
Default image resolution: 940x960
Dataset size: 42240 images
Features:
- brand
- name (not always unique, e.g. Black Leather Loafers)
- gender
- categories
- images
- seoKeyword
- priceByCountry
- url
Service description: Telegram bot with comfortable UX for user
Stack: TBD
├── LICENSE
├── Makefile <- Makefile with commands like `make data` or `make train`
├── README.md <- The top-level README for developers using this project.
├── data
│ ├── external <- Data from third party sources.
│ ├── interim <- Intermediate data that has been transformed.
│ ├── processed <- The final, canonical data sets for modeling.
│ └── raw <- The original, immutable data dump.
│
├── docs <- A default Sphinx project; see sphinx-doc.org for details
│
├── models <- Trained and serialized models, model predictions, or model summaries
│
├── notebooks <- Jupyter notebooks. Naming convention is a number (for ordering),
│ the creator's initials, and a short `-` delimited description, e.g.
│ `1.0-jqp-initial-data-exploration`.
│
├── references <- Data dictionaries, manuals, and all other explanatory materials.
│
├── reports <- Generated analysis as HTML, PDF, LaTeX, etc.
│ └── figures <- Generated graphics and figures to be used in reporting
│
├── requirements.txt <- The requirements file for reproducing the analysis environment, e.g.
│ generated with `pip freeze > requirements.txt`
│
├── setup.py <- makes project pip installable (pip install -e .) so src can be imported
├── sneakers <- Source code for use in this project.
│ ├── __init__.py <- Makes src a Python module
│ │
│ ├── data <- Scripts to download or generate data
│ │ ├── make_dataset.py
│ │ ├── curl_params.py
| | └── ssense_scrapings.ipynb
│ │
│ ├── features <- Scripts to turn raw data into features for modeling
│ │ └── build_features.py
│ │
│ ├── models <- Scripts to train models and then use trained models to make
│ │ │ predictions
│ │ ├── predict_model.py
│ │ └── train_model.py
│ │
│ └── visualization <- Scripts to create exploratory and results oriented visualizations
│ └── visualize.py
│
└── tox.ini <- tox file with settings for running tox; see tox.readthedocs.io
Project based on the cookiecutter data science project template. #cookiecutterdatascience