I missed some time on this project because of travel for work, but I'm back and ready to proceed....
There are several bird databases available for training the ML model so I'm not going to try and make my own.
These are some of the ones I found online and some white papers as examples. There were some databases that were audio characteristic based, and some that were visual. I'm going to focus on the visual data.
Bird image dataset
https://www.kaggle.com/datasets/klu2000030172/birds-image-dataset
Bird Image Dataset Classification using Deep Convolutional Neural Network Algorithm
2022 International Conference on Data Science and Intelligent Computing (ICDSIC)
IEEE Xplore
https://ieeexplore.ieee.org/document/10075724
Identification and Classification of Singing Birds Using Machine Learning
Proceedings of 27th International Symposium on Frontiers of Research in Speech and Music
https://github.com/Moddy2024/Bird-Classification
Birdsnap is a large bird dataset consisting of 49,829 images from 500 bird species with 47,386 images used for training and 2,443 images used for testing.
https://paperswithcode.com/dataset/birdsnap
Building a Convolutional Neural Network to Classify Birds
https://medium.com/@jonschwartz_58395/building-a-convolutional-neural-network-to-classify-birds-528794240fa1
https://github.com/jvk1chaitanya/Birds-Species-Classification
Birds Classification With Pre-trained Convolutional Neural Networks and CAM Heatmap Visualisation
https://medium.com/@bobbycxy/birds-classification-with-pre-trained-convolutional-neural-networks-and-cam-heatmap-visualization-3d424423aaa4
https://github.com/bobbycxy/image-classifier-with-cnn/blob/main/birds-species_training.ipynb
I found one with the birds most likely to visit my location and downloaded the images. I expect hummingbirds, cardinals, blue jays, chickadees, woodpeckers, and such. There are 2750 training images, and 250 images for model testing in the database I downloaded.
I'm going to start by using Jupyter Lab and TensorFlow with this bird characterization/classification database, and then apply the image captured from the webcam attached to the Arty-Z7 to the model.
Top Comments