IoT‐Based Animal Watch Safety Using Green Technology with Deep Learning Approach

In this study, convolutional neural network (CNN)–based method for Internet of Things (IoT)–based object detection in photos of firearms is presented in green technology. The goal of the study is to create a powerful model that can locate and identify weapons in photos with accuracy in IoT. The suggested methodology involves model architecture creation, image preprocessing methods, and training on a dataset of firearm images in green technology. The results show that the CNN-based IoT model is effective at finding and locating weapons with a high degree of accuracy of 96.56%. The preparation of the dataset, which comprises gun images and related label masks, is the first step in the investigation. TensorFlow's image loading routines are used to load the photos, and they are then transformed into NumPy arrays. By giving suitable values to areas where guns are present, label masks are created. These preprocessing steps guarantee that the dataset is suitably ready for the CNN model's training in green technology. The TensorFlow Keras package is used to create CNN architecture for the IoT-based object detection model. Multiple convolutional and pooling layers are used by the model to extract pertinent characteristics from the input photos. Upsampling layers are used to increase spatial resolution; dropout layers are used to prevent overfitting. The network's final layer uses a sigmoid activation function for binary classification to determine whether there are firearms in the photos. The weighted binary cross-entropy loss function and the Adam optimizer are used to train the model. The weighted loss function addresses the class imbalance problem frequently seen in object detection tasks by giving the positive class a higher weight. With a predetermined batch size and validation split, the model is trained iteratively across numerous epochs. The model learns to reliably detect guns in the photographs as training progresses. Predicting the masks for a random sample of photos from the dataset is how the trained model is assessed. The actual photos and the expected masks are then combined to create masked images that emphasize the gun locations. This assessment offers visual proof of the model's capability to locate weapons in various photos. The experimental findings show how reliable and accurate the suggested CNN model for detecting guns is. The model's high accuracy rate of 96.56% shows that it can successfully recognize weapons in a variety of visual settings. The research publication emphasizes that the author does not claim ownership of the photos and notes that the dataset utilized in the study was obtained from Google photos. In this study, CNNs are successfully used to identify objects in photos of guns in green technology. The suggested approach demonstrates the created model's efficacy in precisely recognizing and localizing weapons. The achieved high accuracy rate emphasizes CNNs’ ability to address significant societal concerns about the detection of guns in green technology. This study can be expanded upon to include real-time gun detection systems, which would improve security and safety measures for the general population.