In the field of IoT, object detection can be hampered by factors such as low-light and physical occlusions. Mukherjee et al. address these challenges with “YOLO Phantom,” a lightweight YOLO model featuring a novel Phantom Convolution block, as outlined in their paper MODIPHY: Multimodal Obscured Detection for IoT using PHantom Convolution-Enabled Faster YOLO.

Key characteristics of YOLO Phantom include:

• Notably reduced parameters, model size, and computational intensity compared to YOLOv8n.
• Utilization of transfer learning on a multimodal dataset for enhanced object detection in challenging environments.
• Verification of real-time efficacy on an IoT platform equipped with advanced cameras and AWS-based notification systems.

The research showcases a significant technology advancement for IoT applications requiring accurate, real-time detection, particularly in difficult operational conditions.