Chosen808/insect-identification

Identifying insects from images in the wild using CV and ML

insect-identification

Insect identification from user images in the wild. Built with Python, PyTorch, Weights and Biases. First Class Honours dissertation project.

IF YOU WANT TO USE THE COLAB NOTEBOOKS, MAKE A COPY! DON'T REQUEST ACCESS TO MINE. lol

In Colab: File -> Save a copy to Drive.

Colab Notebooks

• ResNet training over Leeds butterfly dataset
• ResNet training over Leeds butterfly dataset V2
• ResNet inference over Leeds butterfly dataset
• ResNet training with the British Carabid Collection (7 classes version)
• Training with the full IP102
• Training with the full British Carabid Collection
• Inference over 7-class Carabid dataset
• Inference over entire Carabid set
• iNat2021_mini training

---

Gradio demo notebooks

• Butterfly demo code

---

Scripts

• Useful Python scripts created for the project

---

Datasets

• IP102 ¹ (Classification: > 75,000 images in 102 classes, Detection: 19,000 images) --> The dataset can be downloaded from this [link]
• Leeds Butterfly Dataset (Classification: 832 images in 10 classes)
• British Carabid Collection ² - (Classification: > 60k specimens in 291 classes)
• iNaturalist2021 insect subset (Classification: 688,942 insect images in 2,526 classes). available through torchvision.

---

Footnotes

1. Wu, Xiaoping, Chi Zhan, Yu-Kun Lai, Ming-Ming Cheng, and Jufeng Yang. "Ip102: A large-scale benchmark dataset for insect pest recognition." In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 8787-8796. 2019. ↩

2. Hansen, Oskar Liset Pryds, Svenning, Jens-Christian, Olsen, Kent, Dupont, Steen, Garner, Beulhah H., Iosifidis, Alexandros, Price, Benjamin W., & Høye, Toke T. (2019). Image data used for publication "Species-level image classification with convolutional neural network enable insect identification from habitus images " [Data set]. Zenodo. https://doi.org/10.5281/zenodo.3549369 ↩