Evaluation
Achievements
Achievement Table |
||||
|---|---|---|---|---|
| ID | Requirements | Priority | State | Contributors |
| R1 | Apply multimodal RAG for semantic search with text | MUST | ALL | |
| R2 | Apply multimodal RAG for similarity search with image | MUST | ALL | |
| R3 | Store and retrieve text as embeddings in vector database | MUST | ALL | |
| R4 | Store and retrieve images as embeddings in vector database | MUST | ALL | |
| R5 | Chatbot interface with multimodal 'upload to search with' function | MUST | ALL | |
| R6 | Interface for uploading files to database | MUST | ALL | |
| R7 | Integrate chatbot LLM capable of answering text questions | MUST | ALL | |
| R8 | Integrate chatbot LLM capable of analysing images | MUST | ALL | |
| R9 | Map view with locations of stored data | SHOULD | ALL | |
| R10 | Store and retrieve audio as embeddings in vector database | SHOULD | ALL | |
| R11 | Store location of uploaded data | SHOULD | ALL | |
| R12 | Able to display retrieved audio files | SHOULD | ALL | |
| R13 | Kanban-style interface supports task prioritization and custom tags | SHOULD | ALL | |
| R14 | Event geotagging for reports | SHOULD | ALL | |
| R15 | Able to display retrieved video files | COULD | ALL | |
| R16 | Store and retrieve video as embeddings in vector database | COULD | ALL | |
| R17 | Apply multimodal RAG for similarity search with audio | COULD | ALL | |
| R18 | Integrate chatbot LLM capable of analysing audio | COULD | ALL | |
| R19 | Apply multimodal RAG for similarity search with video | COULD | ALL | |
| R20 | Integrate chatbot LLM capable of analysing video | COULD | ALL | |
| R21 | Measure carbon footprint of the system | COULD |
Key functionalities: 100% Completed
Optional Functionalities: 87.5% Completed
Known Bugs |
||
|---|---|---|
| ID | Bug Description | Priority |
| 1 | When the user keeps adding new chats, the chat history session may overflow out of the screen | Low |
Individual Contribution Table For Website |
||||
|---|---|---|---|---|
| Work Packages | Hao | Jack | Penghui | Ruizhe |
| Website Tempate and Setup | 25 | 25 | 10 | 40 |
| Home | 25 | 25 | 10 | 40 |
| Video | 30 | 30 | 0 | 40 |
| Requirement | 30 | 20 | 20 | 30 |
| Research | 5 | 70 | 10 | 15 |
| UI Design | 15 | 15 | 10 | 60 |
| System Design | 85 | 5 | 5 | 5 |
| Implementation | 85 | 5 | 5 | 5 |
| Testing | 5 | 85 | 5 | 5 |
| Evaluation and Future Work | 5 | 5 | 5 | 85 |
| User and Deployment Manuals | 70 | 5 | 5 | 20 |
| Legal Issues | 10 | 10 | 20 | 60 |
| Blog and Monthly Video | 25 | 25 | 25 | 25 |
| Overall contribution | 32.0 | 25.0 | 10.0 | 33.0 |
Individual Contribution Table For System Artefacts |
||||
|---|---|---|---|---|
| Work Packages | Hao | Jack | Penghui | Ruizhe |
| Research and Experiments | 50 | 0 | 5 | 45 |
| UI Design | 50 | 0 | 20 | 30 |
| Coding | 70 | 5 | 5 | 20 |
| Testing | 0 | 95 | 5 | 0 |
| Overall contribution | 42.5 | 25 | 8.75 | 23.75 |
Critical Evaluation
User Interface and experience
We created a clean, intuitive interface for easy multimedia uploads, geographic visualization, task management, multimodal chat, and video interaction. Users found the UI visually appealing, appreciating its clear navigation, intuitive features, and effective multimodal retrieval capabilities. However, improvements can still be made in mobile responsiveness and optimizing load times, particularly for multimedia content.
Overall, our UI/UX receives a rating of Good.
Funtionality
Our web application successfully delivers core functionalities, including multimedia data uploads, geographic visualization, task management, multimodal conversational retrieval, and interactive video analysis. All features performed consistently without reported issues during testing and user interactions, demonstrating robust reliability and accuracy across the platform.
We give a Very Good rating to our functionality.
Maintainability
The system is built with a modular and well-structured codebase, making updates and debugging straightforward. Features like clearly separated components, standardized naming, and clean API integration support long-term scalability and ease of maintenance. Future enhancements can be implemented efficiently without disrupting existing functionality.
We give a Good rating to our maintainability.
Stability
The application performed reliably throughout testing, with all core features functioning as expected. Only one minor, low-priority bug was identified—an overflow in the chat history panel when too many sessions are added. This issue does not affect core functionality and can be easily resolved in future updates.
We give a Adequate rating to our stability.
Efficiency
The application performs well overall, with smooth interactions and fast load times for most features such as file uploads and task management. However, retrieval operations—especially when processing large or complex queries—can take noticeably longer. This may affect the user experience in time-sensitive scenarios and could be optimized further.
We give a Good rating to our efficiency.
Compatibility
The application is compatible with major desktop environments and modern browsers. It runs reliably on macOS (including Apple Silicon via osx-arm64) and supports Docker-based deployment for backend services. However, due to limited mobile optimization and the requirement for tools like Conda, Docker, and Node.js, setup and usage are currently best suited for developers using Unix-like systems. Mobile and broader OS support could be improved for wider accessibility.
We give a Good rating to our compatibility.
Project Management
During development, we used a combination of tools to ensure effective communication and coordination. GitHub was used for version control and issue tracking, while WeChat and email facilitated quick discussions and updates among team members. This multi-channel approach kept everyone aligned and enabled efficient collaboration throughout the project.
We give a Very Good rating to our project management.
Future Work
Faster Retrieval and Indexing
Retrieval performance can be improved by optimizing vector indexing, caching frequently accessed queries, and using more efficient backends for real-time inference. This would reduce latency for large or complex multimodal searches.
User Accounts & Access Control
Introducing user authentication and access control would allow personalized experiences, secure data handling, and role-based task assignment, enabling use in real-world community or organizational contexts.
Scalable Deployment & Cloud Integration
Executing the planned strategies with precision and agility. Our team implements solutions while maintaining constant communication and progress updates.
Carbon Footprint Tracking System
To align with sustainability goals, a future enhancement could include a carbon footprint tracking module. This would allow users to estimate the environmental impact of activities or uploads—such as video processing, data storage, or transportation-related tasks. Integrating APIs for carbon emission estimation or allowing manual data input can support environmentally conscious decision-making and reporting.