punyaslokdutta / Payvment-Design

Payvment provides smart parking solutions.

Payvment-Design

Payvment provides smart parking solutions.

Non Funtional Requirements

1. Real-timeness?
2. Strong Consistency & High Confidence score
3. Low latency (How much latency is acceptable?)
4. Availability (99.999%) - Fault tolerance

Scale Estimation

6. Scaling & High Throughput( 100 parking places * 1000) (10^5/10^5 = 1 QPS) which is low

Functional Requirements

1. LP Detection (Bounding box)
2. Acurracy -> [[position, "ABC", confidence]]
3. accessible by api v1/vehicle/{license_plate_no}

DB model

Possible MongoDb Schema

{
  "_id": "ObjectId", // MongoDB automatically generates a unique identifier for each document
  "licensePlate": {
    "plateNumber": "String",
    "confidenceLevel": "Number"
  },
  "region": {
    "country": "String",
    "stateProvince": "String"
  },
  "vehicleType": "String", // Big Truck, Bus, Motorcycle, Pickup Truck, Sedan, SUV, Van, Unknown
  "makeModel": {
    "make": "String",
    "model": "String"
  },
  "color": "String", // Black, Blue, Brown, Green, Red, Silver, White, Yellow, Unknown
  "directionOfTravel": {
    "degree": "Number",
    "unit": "String" // Degrees, Radians, etc. (based on your preference)
  },
  "dwellTime": "Number", // Time in seconds
  "vehicleOrientation": "String", // Additional information about vehicle orientation if needed
  "timestamp": "Date" // Timestamp of when the record was created
}

ML Models

1. Object Detection -> Yolo V8 Pretrained(Car, Bike , Truck)
2. License Plate detection (Bounding Box)
3. OCR (Easy OCR)

Datasets

1. https://www.kaggle.com/code/mohamedbhy/automatic-number-plate-recognition
2. https://www.kaggle.com/datasets/scholngusmaximus/numberplate-bounding-box-india-eu-brazil-us

Context/Decision Making:

1. Training specific to a number plates region (will lead to higher confidence)
2. Processing on Server vs Edge computing for realtimeness

Edge cases:

Blurry images Low-res images Dark images Small images Vehicles at an angle Vehicles driving fast Vehicles far away Multiple vehicles Plates with icons Plates with stacked characters Plates with 2 rows Plates with tough characters Night-time, reflective plates Motorcycles, buses, trucks Vehicles with signs Vanity plates