---

created: 2025-07-02 tags: [computer-vision] - "#software-architecture" - "#frontend" - "#backend" - database org: - GC goal: Overall design doc of the project

---

Frontend

• More Detailed Plans: 02-Frontend

User Flow - Upload new model

1. Clicks ‘Upload Model’, select a .pt file
2. If success, default select ‘Run Inference’
3. A pop-up window shows up for users to upload image/video
4. Displays results of the image

User Flow - Run existing model

1. User sees the models list.
2. User clicks on the model name to select this model
3. Clicks ‘Run Inference’
4. A pop-up window shows up for users to upload image/video
5. Displays results of the image

Component Tree

![[Screenshot 2025-07-02 at 10.54.04 AM.png]]

HTTP Request

1. Upload model
   1. form-data
      1. .pt file
      2. name
2. Run inference
   1. form data
      1. event_id
      2. model id
      3. temp path to file (image/video)

Backend

API Endpoints

1. Clicks ‘Upload Model’, select a .pt file
   1. POST /api/models
2. User sees their custom model in the models list.
   1. GET /api/models
3. User clicks on the model name to select this model
   1. Frontend just stores the selected model id
4. Clicks ‘Run Inference’
   1. POST /api/run
5. Displays results of the image

GET /api/models

• Returns a list of models in json format
• .json file should include: id, model name, path

POST /api/models

• (User uploads a new .pt file)
• Save the file to a path (backend/app/models/{model-name})
• Update models.json (id, model name, path)

DELETE /api/models/:id

• Update models.json

POST /api/run

• Request body:
  • event_id
  • model id
  • temp path to file (image/video)
• Loads the .pt file for that model
• Gets the file
• Runs the inference
• ~~(Stores the results in DB)~~
• Return results as JSON

Should have two blueprints:

• model
• inference

Database

Fields:

• event_id (PK)
• model_id
• results
• input_file
• timestamp

Deployment

Reference

previous proposal: CV Playground Backend