FYP25043
Overview Product Coverage AI & CV Architecture Evidence

HKU Library Booking System with Real-time Space Monitoring for iOS and Android

A completed mobile platform for HKU students to find study space, book library facilities, view live occupancy, receive reservation reminders, analyze historical crowd patterns, and complete bookings through an AI assistant.

Students: Cheng Chak Yuen, Hao Qiancheng, Luo Siqi
Supervisor: Dr. Zhao Hengshuang
Computer vision occupancy detection result in a study area
Computer vision detection result showing study seats used for real-time occupancy and availability estimation.

Project Snapshot

A complete booking, occupancy, AI, and analytics platform.

The system is implemented as a Flutter mobile app, FastAPI backend, AI booking service, computer-vision module, and local infrastructure stack for database, cache, queues, and video streaming.

5

HKU library locations in the booking catalog.

236

Bookable facilities seeded across the supported libraries.

4

Local camera demo streams for the occupancy pipeline.

6

Core app modules: Home, AI, Libraries, Reservations, Reports, and Settings.

Completed Product

The app supports the full student study-space journey.

Students can move from space discovery to reservation confirmation, manage upcoming bookings, and use live and historical data to choose better study locations.

Account and secure session

Students register, log in with HKU email, keep a stored session, and access authenticated booking, AI, occupancy, and report APIs.

Live occupancy home

The home page uses location and current occupancy snapshots to list nearby study areas and recommend a best spot by distance or crowd level.

Library and facility browsing

Students browse supported HKU libraries, inspect facility groups, open facility maps, and choose a target room, table, booth, or computer.

Timeslot booking

Facility timeslots are loaded from the backend with availability status, then students confirm a reservation from the app.

Reservation management

Users can review all, upcoming, and historical reservations, cancel eligible bookings, and open the relevant facility map from a reservation.

Insights dashboard

Historical reports summarize occupancy, trends, heatmaps, and peak hours to help students avoid crowded study periods.

User Flow

A simple path from finding space to keeping the booking.

1

Check live space

Open the home page and compare nearby library occupancy.

2

Choose a facility

Browse libraries, rooms, tables, booths, computers, and studios.

3

Book a timeslot

Select an available slot and create a confirmed reservation.

4

Get reminded

Use local notification reminders before the booking starts.

5

Plan smarter

Use reports and peak-hour patterns for the next study session.

Campus Coverage

The catalog covers multiple HKU libraries and facility types.

Seed data includes aliases, coordinates, opening hours, capacities, floors, room codes, display names, and bookable facility definitions.

Supported locations

Main Library 111
Discussion rooms, study tables, computers, studio and editing room, concept and creation rooms, and single study rooms
Lui Che Woo Law Library 94
Study tables and discussion rooms
Chi Wah Learning Commons 18
Study rooms and study booths
Yu Chun Keung Medical Library 10
Discussion rooms and single study rooms
Music Library 3
Discussion rooms

Bookable facility types

The system goes beyond discussion rooms and supports a broader set of individual and collaborative study facilities.

Discussion Room
Study Table
Study Room
Study Booth
Single Study Room
Computer in LTC
Studio and Editing Room
Concept and Creation Room

AI and Computer Vision

Smart assistance is built into both booking and space monitoring.

The project combines a conversational booking assistant with a computer-vision occupancy pipeline for real-time study-space awareness.

Conversational booking agent

  • Creates and manages user-specific AI sessions.
  • Collects library, room type, date, time, and room choices.
  • Maps natural-language library names to legacy booking codes.
  • Shows room suggestions and supports map-based room selection.
  • Requires final confirmation before creating a reservation.

Occupancy detection pipeline

  • Uses person/item detection and seat detection together.
  • Counts people as occupied seats.
  • Detects seat hogging when items occupy seats without nearby people.
  • Aggregates camera-derived logs into area-level occupancy snapshots.
  • Supports Redis windows, RabbitMQ queues, and local demo streams.

Architecture

Each subsystem can run independently or as a full local stack.

Flutter app

Home, AI chat, booking, reservations, reports, notifications, and map viewer.

Database and time-series storage

PostgreSQL and TimescaleDB store users, reservations, occupancy logs, snapshots, and reports.

FastAPI backend

Auth, libraries, facilities, reservations, AI orchestration, occupancy, reports, and health checks.

Cache and queue infrastructure

Redis handles live occupancy cache. RabbitMQ stages camera frames and normalized stats events.

AI agent service

Python service for natural-language booking sessions and confirmed reservation execution.

CV and streaming tools

OpenCV, PyTorch, Ultralytics, MediaMTX, and demo streams support occupancy testing.

Technology Stack

The implementation stack used in the completed project.

Mobile

Flutter, Dart, Provider, secure storage, local notifications, geolocator, permission handler, and Google Fonts.

Backend

FastAPI, SQLAlchemy, JWT authentication, APScheduler, modular routers, services, and OpenAPI documentation.

Data

PostgreSQL, TimescaleDB, occupancy logs, area snapshots, rollups, retention policies, and seed data scripts.

Infrastructure

Docker Compose, Redis, RabbitMQ, MediaMTX, RTSP, RTMP, WebRTC, and local video fixtures.

AI

Python AI service, OpenAI-compatible LLM API, session management, booking confirmation, and backend orchestration.

Computer vision

OpenCV, PyTorch, Ultralytics YOLO models, custom seat/person/item detection, training scripts, and test images.

Testing and Evidence

The project includes working code, fixtures, tests, and operational documentation.

Frontend tests

Repository, controller, provider, model parsing, network, notification, and widget tests are included.

API documentation

The backend exposes Swagger UI, ReDoc, health checks, and module-level API documentation.

Local full-stack setup

Docker Compose and environment examples support PostgreSQL, Redis, RabbitMQ, MediaMTX, backend, AI, and Flutter bring-up.

Deliverables

What this project delivers.

Mobile application

Cross-platform Flutter app for iOS and Android.

Backend API

Authenticated FastAPI service for booking, occupancy, AI, and reports.

AI agent

Natural-language booking assistant connected to the reservation system.

CV pipeline

Seat and occupancy detection with camera pipeline support.

FYP25043 Site

© 2026 HKU Library Booking System with Real-time Space Monitoring for iOS and Android