/CV-InceptionNN

Classification of 120 breeds with DNN and inception blocks

Primary LanguageJupyter Notebook

Dogs breed classification

This dataset contains images of 120 breeds of dogs from around the world. Total number of images is 20,580.

The prediction model consist of pre-trained MobileNetV3 model, 2 extra inception blocks and 1 short connection. That's how the model looks like:

Model scheme

The model is saved in best_model.h5 file.
Test accuracy of the model is 0.946, top 2 accuracy is 0.9764