APAP-AI

This is a repository for an object detection inference API using YOLOv8 and FastAPI

This repository is used for 스마트해상물류 ICT project.

How to Use

Operations

There are three operations: estimate_distance, custom_model, and area_intrusion. Each of these operations can be used to detect different anomalies in video.

Operation	Description
estimate_distance	GitHub Container Registry Token
custom_model	Remote server IP
area_intrusion	Remote server user name

PUB-SUB

In this system, a publisher (camera) sends frames to multiple subscribers (users) via WebSocket. Subscribers receive processed frames based on the operation specified. For example:

Publisher

{host}/ws/publishers/{location_name}?op={operation_name}

• {host}: APAP AI server address
• {location_name}: Camera location or context
• {operation_name}: Operation to apply (e.g., estimate_distance)

Subscriber

{host}/ws/subscribers/{location_name}

• {host}: APAP AI server address
• {location_name}: Camera location or context

---

How to Run

Install

Dependencies are managed using Poetry.

 pip install poetry
 poetry install

Run

poetry run uvicorn --host=127.0.0.1 app.main:app

Usage

Environment variables are managed through dotenv.

The defined variables are referenced from the .env file in the root directory.

API Endpoints

Documentation is provided through SwaggerUI, which is built into FastAPI by default.

 http://localhost:8080/docs

Test

 poetry run pytest

Deployment via GitHub Actions

The following variables need to be defined in the Environment secrets:

Variable	Description
GHCR_TOKEN	GitHub Container Registry Token
REMOTE_IP	Remote server IP
REMOTE_USER	Remote server user name
REMOTE_PRIVATE_KEY	Remote server private key
REMOTE_SSH_PORT	22