DEVELOPMENT OF A BANDWIDTH ALLOCATOR USING THE FASTER R-CNN ALGORITHM FOR VIDEO ANALYTICS APPLICATIONS (VAPS)

This digital product presents a comprehensive and technical exploration of video analytics optimization using the Faster R-CNN object detection algorithm in bandwidth-constrained environments. Tailored for smart city applications such as traffic control and crowd monitoring, this product includes a detailed research presentation and supporting materials developed as part of a doctoral program at Institut Teknologi Bandung.

Key Highlights:

• ✅ Explains the architecture of Faster R-CNN and its use in object detection tasks in real-time video analytics.
• 📹 Analyzes Video Analytics Applications (VAPs) across smart cities, public safety, retail, and healthcare.
• ⚙️ Presents a proposed bandwidth allocator model to optimize video transmission by dynamically adjusting compression parameters (Quantization Parameter and FPS).
• 📊 Experimental evaluation showcasing trade-offs between bitrate, F1-Score, and PSNR using H.264 codec.
• 🧠 Machine learning integration with k-Nearest Neighbors for classifying optimal video encoding settings.
• 📈 Supports both objective and subjective video quality assessments, including MOS and JND.
• 💡 Includes a novel resource allocator framework comparison: Static vs Dynamic allocation mechanisms.

Perfect for:

• Researchers in AI and video analytics
• Network engineers and smart city planners
• ML developers focusing on video compression
• Students and academics exploring resource allocation techniques

🔗 Bonus: Includes quiz access and references to major IEEE papers and open-source repositories.