Birds_Classifier

Dataset

Data set of 150 bird species. 19744 training images, 750 test images(5 per species) and 750 validation images(5 per species.All images are 224 X 224 X 3 color images in jpg format. Also includes a "consolidated" image set that combines the training, test and validation images into a single data set. source: https://www.kaggle.com/gpiosenka/100-bird-species

Model Architecture

Inspired by GoogLeNet/Inception v1, has a total of 10 million trainable parameters and was trained for only 30 epochs. Refer the uploaded jupyter notebook for details about the modle architectue.

Model file

Model has been converted json format for ease of use, File name: "model.json"

Trained Weights

Training achieved an accuracy of 84.67 percent on the validation set. File name: "model_weights_v3.h5"