Code for Butterfly and Moth Species Detection and Classification Using Deep Learning

Butterflies bring to mind images of meadows, colorful flower filled landscapes
and lively summer gardens. Aside from their appeal, areas rich in butterflies and
moths indicate an ecosystem that supports a wide variety of invertebrate populations. This research delves into the classification of images specifically focusing on
identifying 100 species of butterflies and moths. The dataset used consists of 12,594
training images 500 validation images and 500 test images all sized at 224 x 224
pixels. By employing four models including pretrained models such as EfficientNetB0, ResNet50, VGG19 Models and a custom CNN model named ButterflyNet.
ResNet50 stands out with a test accuracy of 95% closely followed by EfficientNetB0 at 93.60% VGG19 Model at 92.80% and ButterflyNet at 85.40%. Moreover
incorporating an interactive Streamlit UI enhances accessibility by allowing users
to conduct real time tests. In conclusion ResNet50 emerges as the model while
ButterflyNet shows promising potential. Future efforts should explore tuning techniques, ensemble methods and continuous model optimization to contribute to the
evolving field of image classification and its crucial role, in biodiversity conservation
through technological advancements.

Скачать видео