System Requirements Document (SRD)

Project Title: eagle-ai

1. Introduction

Welcome to the System Requirements Document (SRD) for eagle-ai, an advanced AI-powered system designed to deliver cutting-edge object detection capabilities. This document outlines the functional and non-functional requirements, design concepts, and technical specifications necessary to build and deploy the system successfully.

eagle-ai leverages Roboflow's RF-DETR 2 model for high-accuracy object detection and integrates a visually dynamic, 3D-inspired user interface to ensure seamless interaction and real-time performance. The system is designed for deployment online, ensuring accessibility and scalability for users across India and beyond.

2. System Overview

eagle-ai is an AI-driven object detection platform that processes input images or webcam frames through Roboflow's RF-DETR 2 model. The system pipeline is as follows:

1. Input: Images or webcam frames are uploaded or captured.
2. Processing: The input is sent to Roboflow's hosted inference API or a local inference server.
3. Detection: RF-DETR 2 processes the input and returns predictions, including bounding boxes, class labels, and confidence scores.
4. Output: Results are rendered visually on the user interface using advanced annotation tools.

The system is optimized for deployment online, ensuring scalability and accessibility for users. It features a premium, glassmorphic UI with 3D elements, making it visually engaging and intuitive.

3. Functional Requirements

As story points:

• As a User, I should be able to upload images for object detection.
• As a User, I should be able to start live webcam detection.
• As a User, I should be able to view detection results with bounding boxes, class labels, and confidence scores.
• As a User, I should be able to adjust detection confidence thresholds using a slider.
• As a User, I should be able to select different model sizes (nano → xlarge).
• As a User, I should be able to toggle detection ON/OFF.
• As a User, I should be able to view a dashboard displaying FPS, objects detected, and average confidence.
• As a User, I should be able to interact with a visually dynamic 3D background on the homepage.

4. User Personas

1. Admin

• Responsible for managing system configurations and monitoring performance.
• Can access advanced settings for inference modes and API keys.

2. User

• Primary end-user who uploads images or uses the webcam for object detection.
• Interacts with the UI to view detection results and adjust settings.

3. Guest

• Limited access to explore the system's features without uploading images or using the webcam.

5. Visuals Colors and Theme

Color Palette:

• Background: #0f172a (Dark theme)
• Surface: #1e293b (Glassmorphic blur cards)
• Text: #e2e8f0 (Light gray for readability)
• Accent: #38bdf8 (Neon cyan glow)
• Muted Tones: #64748b (Subtle gray for secondary elements)

Theme:

The design follows a glassmorphic aesthetic with neon glow accents and smooth animations. Rounded corners and a floating UI feel enhance the user experience.

6. Signature Design Concept

Interactive 3D Galaxy Homepage

The homepage of eagle-ai will feature an interactive 3D galaxy map created using Spline or Three.js. Users will navigate through a futuristic galaxy where each star represents a feature or section of the platform.

Details:

• Background: A dynamic galaxy scene with glowing stars and nebulae.
• Interaction: Hovering over stars reveals tooltips with feature descriptions.
• Animations: Smooth transitions as users move between sections.
• Micro-interactions: Clicking on a star zooms into the corresponding section, such as "Upload Image" or "Live Detection."
• Color Shifts: The galaxy subtly changes colors based on user interactions, transitioning between cyan, purple, and muted tones.

This design ensures a memorable first impression and aligns with the futuristic theme of the project.

7. Non-Functional Requirements

• Performance: The system must process images and webcam frames in real-time with minimal latency.
• Scalability: The system must support multiple concurrent users without degradation in performance.
• Security: API keys and sensitive data must be securely stored and accessed.
• Accessibility: The UI must be intuitive and accessible to users of varying technical expertise.
• Deployment: The system must be deployed online, ensuring global accessibility.

8. Tech Stack

Frontend:

• React for Web
• Streamlit for rapid UI development

Backend:

• Python
• FastAPI for API integration

Database:

• MySQL for structured data storage
• MongoDB for unstructured data storage

AI Models:

• RF-DETR 2 for object detection
• GPT 5.4 for user-friendly responses

AI Tools:

• Litellm for LLM Routing
• Langchain for pipeline orchestration

Local Orchestration:

• Docker
• docker-compose

Server-Side Orchestration:

• Kubernetes for deployment and scaling

9. Assumptions and Constraints

Assumptions:

• Users will primarily access the system online.
• Roboflow's hosted inference API will be used for object detection.
• The system will support real-time detection with minimal latency.

Constraints:

• Deployment must adhere to cloud-based requirements.
• The system must operate within the limits of Roboflow's API usage policies.
• The UI must be optimized for performance without compromising visual quality.

10. Glossary

• RF-DETR 2: A transformer-based object detection model developed by Roboflow.
• Glassmorphism: A UI design trend characterized by translucent surfaces and blur effects.
• Spline: A tool for creating 3D scenes and animations.
• Inference: The process of running a trained AI model to make predictions on input data.
• Bounding Box: A rectangular box used to highlight detected objects in an image.
• Confidence Score: A numerical value indicating the certainty of a detection.

Let me know if you'd like further refinements or additional sections, Yashasvi!