Development of a deep learning model from breeding substrate images: a novel method for estimating the abundance of house fly (Musca domestica L.) larvae

An annotated image dataset for training mosquito species recognition system on human skin

This paper introduces a new mosquito images dataset that is suitable for training and evaluating a recognition system on mosquitoes in normal or smashed conditions. The images dataset served mainly for the development a machine learning model that can recognize the mosquito in the public community, which commonly found in the smashed/damaged form by human. Especially the images of mosquito in hashed condition, which to the best of our knowledge, a dataset that fulfilled such condition is not available. There are three mosquito species in the dataset, which are Aedes aegypti, Aedes albopictus and Culex quinquefasciatus, and the images were annotated until species level due to the specimen was purely bred in a WHO accredited breeding laboratory. The dataset consists of seven root files, six root files that composed of six classes (each species with either normal landing, or random damaged conditions) with a total of 1500 images, and one pre-processed file which consists of a train, test and prediction set, respectively for model construction.

Measurement(s)	recognition system with machine/deep learning
Technology Type(s)	Camera Device
Sample Characteristic - Organism	Aedes aegypti • Aedes albopictus

Classification of Zophobas morio and Tenebrio molitor using transfer learning

Zophobas Morio and Tenebrio Molitor are popular larvae as feed ingredients that are widely used by animal lovers to feed reptiles, songbirds, and other poultry. These two larvae share a similar appearance, however; the nutritional ingredients are significantly different. Zophobas Morio is more nutritious and has a higher economic value compared to Tenebrio Molitor. Due to limited knowledge, many animal lovers find it difficult to distinguish between the two. This study aims to build a machine learning model that is able to distinguish between the two. The model is trained using images that are taken from a standard camera on a mobile phone. The training is carried on using a deep learning algorithm, by adopting an architecture through transfer learning, namely VGG-19 and Inception v3. The experimental results on the datasets show that the accuracy rates of the model are 94.219% and 96.875%, respectively. The results are quite promising for practical use and can be improved for future works.