FIGURE 3.

The CNN was trained on captured mosquito images and feature points through convolution. However, with increasing pixel size of the image, number of feature pixels, and number of features found, the computational cost dramatically increases. Therefore, the pooling method was used to compress the images while retaining essential image features to increase performance. Finally, one hot encoding classification was performed to differentiate Ae. aegypti and Cx. quinquefasciatus.