eagle-ai

byYashasvi paheriya

Act as a senior AI engineer, MLOps engineer, and 3D UI/UX designer. You are building a COMPLETE AI system called "VisionX". ⚠️ IMPORTANT CONTEXT: This project uses RF-DETR 2 from Roboflow for object detection. You MUST understand this pipeline: - Input image / webcam frame - Sent to Roboflow Inference API (or local inference server) - Model (RF-DETR 2) processes image - Returns predictions (bounding boxes, class labels, confidence) - Results are rendered visually on UI Your job is to build BOTH: 1. Frontend (Streamlit UI with 3D feel) 2. Proper integration with this AI pipeline -------------------------------------------------- 🔐 ENV VARIABLES (USE THESE) -------------------------------------------------- ROBOFLOW_API_KEY INFERENCE_MODE = "api" or "local" LOCAL_INFERENCE_URL = http://localhost:9001 HOSTED_INFERENCE_URL = https://serverless.roboflow.com -------------------------------------------------- 🧠 MODEL WORKFLOW (MUST IMPLEMENT) -------------------------------------------------- - Use inference-sdk (InferenceHTTPClient) - Dynamically choose endpoint: if INFERENCE_MODE == "local": use LOCAL_INFERENCE_URL else: use HOSTED_INFERENCE_URL - Send image to model: client.infer(image, model_id="rfdetr-medium") - Parse response: Extract: - bounding boxes - class names - confidence - Render using supervision: - BoxAnnotator - LabelAnnotator -------------------------------------------------- 🎨 UI DESIGN (PREMIUM) -------------------------------------------------- - Dark theme (#0f172a) - Glassmorphism (blur cards) - Neon glow accents (cyan/purple) - Smooth animations - Floating UI feel - Rounded corners -------------------------------------------------- 🌐 3D BACKGROUND (MANDATORY) -------------------------------------------------- Embed Spline or Three.js scene using: st.components.v1.html() -------------------------------------------------- 📐 UI STRUCTURE -------------------------------------------------- 1. HERO SECTION - VisionX 🚀 - AI-Powered Object Detection - Description: "VisionX is powered by RF-DETR 2 from Roboflow, delivering high-accuracy object detection using transformer-based architecture." - Add: "Built using Roboflow RF-DETR for state-of-the-art accuracy and real-time performance." -------------------------------------------------- 2. ABOUT SECTION - Explain system pipeline clearly: Input → Model → Detection → Output - Feature Cards: - Real-Time Detection - Transformer-Based Accuracy - Scalable AI Deployment -------------------------------------------------- 3. UPLOAD + DETECTION SECTION - Drag & Drop upload - Process image using model - Show: LEFT → Original image RIGHT → Annotated detection -------------------------------------------------- 4. LIVE DETECTION - Button: Start Webcam - Show placeholder or instructions -------------------------------------------------- 5. DASHBOARD - Show: - FPS (static or simulated if needed) - Objects detected - Avg confidence -------------------------------------------------- 6. FOOTER - "Built with AI • VisionX • Powered by Roboflow RF-DETR" -------------------------------------------------- ⚙️ SIDEBAR CONTROLS -------------------------------------------------- - Confidence slider - Model selector (nano → xlarge) - Toggle detection ON/OFF -------------------------------------------------- ✨ UI ENHANCEMENTS -------------------------------------------------- - Gradient animated text - Glow buttons - Hover scale effects - Blur glass cards -------------------------------------------------- 🧱 CODE STRUCTURE -------------------------------------------------- Functions: - render_hero() - render_about() - render_upload() - run_inference() - render_dashboard() - render_footer() -------------------------------------------------- 📦 OUTPUT FORMAT -------------------------------------------------- - Full working Streamlit code - Clean, modular, production-ready - No missing parts -------------------------------------------------- 🎯 GOAL -------------------------------------------------- Build a real AI product UI that understands the full ML pipeline — not just a frontend mock.