Week #2 #
1. Detailed Requirements Elaboration #
Project Goal #
Build an AI-powered fashion assistant that helps users:
Choose clothing items based on their photo, figure, and style preferences
Assemble capsule wardrobes and plan outfits
See visual try-ons via 3D model
Scollect images from real products from vb
Outfit Recommendation Engine #
Input:
Color type (from photo)
Body shape (from photo)
Optional user style query (“I want something romantic for spring”)
Optional constraints: max price, preferred color, fabric, store
Processing Logic:
Match the user profile to compatible items scraped from WB
Apply filters (price, size, color, category)
Output:
Visual list of 5–10 recommended items
Ability to “save to favorites” or “add to capsule”
UX Notes:
Each item shows image, size availability, price, store, and rating
Visual badges: “Matches your figure”, “In your color palette”, etc.
Capsule Wardrobe Generator #
Input:
User-selected styles OR system-generated suggestions
Number of looks or days (e.g., “capsule for 5 days”)
Preferred color palette and categories (tops, bottoms, outerwear)
Logic:
Combine compatible items into 3–7 interchangeable outfits
Use basic fashion rules: color harmony, proportions, layering
Output:
- Saved capsule with preview (grid of outfits)
UX Notes:
Drag-and-drop feature for manual capsule editing
Capsule preview = grid layout (Mon–Fr)
3D Try-On Preview #
This is a very popular feature but within 7 weeks we plan to make a 2D avatar so that you can try on clothes and see the compatibility
the idea is to make the application more interesting for the user and lay the foundation for a future 3D model
Scraping Wildberries (WB) #
Scrape periodically:
Title, price, photos, size availability, brand, composition
User reviews and user-uploaded photos
Store enriched dataset with parsed tags (e.g., boho, summer, organic cotton)
Authentication & DB #
User login via email + password (JWT-based auth)
DB tables: Users, Favorites, Capsules, Looks, BodyProfile scheme
Constraints & Challenges #
WB HTML structure may change — scraper must be modular and resilient
Some fashion rules (like layering logic or trend-matching) may require manual curation for now
2. Prioritized backlog #
3. User flow diagram #
4. Updated / Detailed User Stories with Acceptance Criteria #
1. Set a Maximum Price #
User Story: As a budget-conscious user, I want to set a price limit so I don’t see unaffordable options.
Acceptance Criteria:
User can define a price cap
All items shown are below that value
Price clearly visible on each card
2. Show Only Available Sizes #
User Story: As a user, I want to see only items available in my size so I don’t get disappointed.
Acceptance Criteria:
User sets size once
Results show only items in stock for that size
“Out of stock” labels or filters available
3. Suggest Outfits for My Body Type #
User Story: As a user, I want outfit suggestions that suit my body type so I feel confident.
Acceptance Criteria:
Personalized tips per figure
Recommended styles for each type
Visual examples provided
4. Color Coordination Assistant #
User Story: As a user, I want help matching colors so my outfits look harmonious.
Acceptance Criteria:
User can choose base color
App suggests matching combinations
Example looks with selected palette
5. Jeans/Trousers for Petite Height #
User Story: As a petite user, I want trousers that fit me in length so I don’t look awkward.
Acceptance Criteria:
Filter/tag for petite sizing
Inseam info shown
Height recommendation on item card
6. International Size Conversion #
User Story: As a user, I want to see size conversions across countries and brands.
Acceptance Criteria:
RU ⇄ EU ⇄ US ⇄ Asia chart
Store-specific sizing guidelines
Auto-convert based on user profile
7. 3D Try-On #
User Story: As a user, I want to preview clothes on a 3D model of myself to check fit and look.
Acceptance Criteria:
Avatar created from body parameters
Clothes visually “worn” by avatar
Can rotate and zoom in viewer
8. Capsule Wardrobe Builder #
User Story: As a user, I want to build a capsule wardrobe to mix and match items easily.
Acceptance Criteria:
Generate 5–10 item capsule sets
Combinations previewed visually
Save/edit capsule looks
9. Weekly Wardrobe Planner #
User Story: As a user, I want to plan my weekly outfits to avoid daily stress.
Acceptance Criteria:
Calendar interface for outfit slots
Drag/drop saved looks
Add notes like “rainy day” or “event”
10. Smart Search with Tags #
User Story: As a user, I want smart search that understands my style keywords.
Acceptance Criteria:
Search supports style terms (e.g., boho, minimalist)
Keyword mapping to tags
Filters adapt to description
11. Boost Self-Esteem #
User Story: As an insecure user, I want positive feedback and suited suggestions so I feel good.
Acceptance Criteria:
Friendly body-type feedback
Focus on what suits, not flaws
No negative language
12. Filter by Country of Origin #
User Story: As a user, I want to choose the country of manufacture to avoid low-trust brands.
Acceptance Criteria:
Show country flag/origin on product card
Filter by origin
Option to exclude certain countries
5. Interviews & anlysis #
6. Project specific progress #
Backend: almost wrote the logic for 2 features that will be included in mvp
Design: finished developing the design, held several meetings on this topic, discussed with the user
Frontend: started writing 1 of 4 pages that will be needed for mvp
ML: connected llm but it requires revision, created a simple model for determining the color, but have not connected it yet
Deployment: added to the docker file launch of llm and ml service, but due to ollama it does not work yet - we are launching them separately
CMO: agreed on advertising with a blogger in tg (audience of 15 thousand on instagram), after creating mvp or closer to the end of development we will conduct
Public relation&CustDev: conducted a survey, 60+ people took part, conducted several personal interviews - results at the link
7. Individual contribution of each participant #
- Ksenia Korchagina (lead) - conducting design meetings, design review, questionnaire review, compiling survey analysis, writing a report
- Yasmina Mamadalieva - training a model to determine color type, conducting interviews with young mothers
- Aisylu Fattakhova - connecting LLM, writing the frontend, integrating the backend service with the LLM service, conducting interviews
- Ekaterina Akimenko - completed the design for the pages that will be included in the MVP, compiled a survey
- Sofia Goryunova - trained the model for the color type, improved its quality, compiled a survey and actively distributed it
- Alena Starikova - almost finished writing the backend functions for mvp, conducted interviews
- Rokkel Maria - developed the design, started writing the frontend, conducted interviews
8. Plan for Next Week #
[23] FRONT transition to feature pages (w3)
[24] FRONT connect logic for 2 features with the back (w3)
[25] FRONT handle errors (w3)
[26] BACK test and finish connecting features
[27] BACK Upload everything to the main branch (w3)
[28] ML catch errors (w3)
[29] ML finish training the mode (w3)
[30] ML prepare a dataset for display (w3)
[31] ML Upload everything to the main branch (w3)
[32] FRONT Upload everything to the main branch (w3)