Integrating Automated Object Recognition Systems in Daycare Environments: Challenges and Solutions Using Machine Learning

This paper introduces a hybrid approach that combines convolutional and recurrent neural networks (CNN-RNN) with temporal attention are combined for real-time object recognition in daycare settings. The proposed system addresses key challenges in dynamic and privacy-sensitive childcare environments. Data were gathered from various daycare centers and used to train and test the model against state-of-the-art models such as Faster R-CNN, YOLOv5s and YOLOv8n. The experimental results indicate that the overall accuracy of the hybrid CNN-RNN model was the highest, 90.6, and was 0.7 and 3.9 higher than the YOLOv8n and Faster R-CNN, respectively, which is statistically significant (p = 0.013). Temporal attention also enhanced the recognition of subtle changes in behavior particularly when it comes to the discrimination of behaviors such as resting, playing and distress state. The architecture suggested was implemented in the edge devices through pruning and quantization optimization. These results highlight the possibility of applying the model to daycare centers, as well as contributing to the active monitoring of the safety of children and more effective systems of this staff in daycare centers.