Skip to content

Latest commit

 

History

History
11 lines (10 loc) · 2.45 KB

proposal.md

File metadata and controls

11 lines (10 loc) · 2.45 KB
Raw

Project Proposal

  1. Abstract: We will build an outfit recommendation local app. This app will be able to read a photo path of a outfit owned by the user. After recognizing the several information outfits owned by the user, recommend several style matching options for the user to choose by deep learning. This project will also recommend new outfits in user's favorite styles to them through methods such as web scraping and Natural language processing from shopping website and fashion forum.
  2. Planned Deliverables: We plan that the final result will be a local app. Its workflow is: users save photos of their own clothes in a specific local path, and the app can extract these photos with the user's permission, and recognize the type, color, and even brand of the clothes through deep learning. Finally, it will generate some clothes matching according to different styles for users to choose. In a more advanced version, it will obtain the latest information in the fashion industry through web scraping, and recommend new clothes based on the user’s favorite style.
  3. Resources Required: We need a data set. As far as we know, there are some training sets for computer vision recognition of clothes in Kaggle. We will also obtain more data through web scraping (pictures and types, brands, etc.).
  4. Tools/Skills Required: In our plan, this project requires machine learning, computer vision, web scrapping. The packages might be numpy, sklearn, Tensorflow, scrapy, etc.
  5. Risks: If it turns out that the data does not exist, we do not need to change the plan, we will find the data ourselves through web scraping. Our recommendation system may be more complicated than imagined. If we cannot complete a complete recommendation system, we will replace the web app plan with a package plan.
  6. Ethics: We're just providing a portfolio of recommendations. User can choose different styles. And they may or may not adopt our advice. So I thing our project won't hurt anyone.
  7. Tentative Timeline: After 2 weeks(we should build a computer vision model, it can recognize the type and color of a photo of cloth), After 4 weeks(Improve our CV model make it can recognize more than one cloth in a single picture. We could get our training data by put some photo of different style of dressing through this model. Then use this training data fit our outfit suggestion model) After 6 weeks(create a web app use our models)