davidwonghk / alert_task-main

Video events management task

This repo contains a short interview coding task at Coram.AI

You are asked to implement backend APIs of a video event management system supporting basic operations of:

1. Ingestion of new video events.
2. Aggregation of recorded video events.
3. Alert triggered on unusual activity.

The events are stored in postgres database the backend interacts with. In our case backend are simple python functions.

Setup

1. Install docker and docker-compose
2. Run the code using ./run.sh. This script starts the database and runs implemented APIs.

Tasks

All tasks should be implemented within main.py file:

Ingestion

To start off, please complete the ingest_data function - feel free to change the signature if you want to. Please also consider the implications of the code - performance, maintenance, etc.
The events have a timestamp and a type property. The type will always be one of a set of pre-defined strings - "pedestrian", "bicycle", "car", "truck", "van". Assume the events are ordered.

Aggregation

We'd like to show our user an overview of the events in the database. To do that please complete the "aggregate_events" function, so that it returns two lists of periods of activity. One should contain the activity of people who were detected (the "pedestrian" and "bicycle" event types) and the other should contain the activity of vehicles (the "car", "truck" and "van" event types). The goal is to aggregate the events in the database into periods of activity in each category. Any events less than 1 minute apart in the same category should be combined into an interval.

For example if we assume that the system has detected cars at the following timestamps: 18:30:30, 18:31:00, 18:31:30, 18:35:00, 18:35:30, 18:36:00, 18:37:30, 18:38:00. The aggregate function should return three periods of activity: 18:30:30-18:31:30, 18:35:00-18:36:00 and 18:37:30-18:38:00.

{
    "people": [],
    "vehicles": [
        ("2023-08-10T18:30:30", "2023-08-10T18:31:30"),
        ("2023-08-10T18:35:00", "2023-08-10T18:36:00"),
        ("2023-08-10T18:37:30", "2023-08-10T18:38:00"),
    ],
}

Please have the whole aggregation logic in SQL.

Alerts

The last part is related to real-time alerts for unusual events. Our users would like to be notified if a person is detected for a continued period of time. Imagine that the ingestion function is called real-time as the detections happen every 30 seconds.

Please add additional logic to the ingestion function which will print to the console if a person is detected in 5 consecutive events. Please consider performance - the ingestion function will be called very often.