Real-time classification of haze and non-haze images on Arduino Nano BLE using Edge Impulse

Image dehazing is of utmost research interest and has several significant applications. However, the technology of image dehazing can only work for hazy images. Therefore, classification of images into haze and non-haze becomes a very essential precursor for the technology of image dehazing. Relying on human visual perception to classify the images is often trivial, empirical, and unrealistic. As a result, there is a requirement for a suitable algorithm that can perform this classification efficiently. This paper aims to meet such requirement and classifies images presented as ‘clear’ or ‘hazy’. Arduino Nano 33 BLE coupled with a Tiny Machine Learning Shield has been used for the implementation of the algorithm. Edge Impulse has been used for training, testing and validation of the Convolutional Neural Network (CNN) model. The image to be classified is displayed to the camera in real-time. Subsequently, the captured image is given to the model, which then classifies the image. The output of the classification is displayed on the serial monitor.