Dogs vs cats

Purpose: Using convolutional neural network technology to identify whether the animal in the target photo is a cat or a dog.

Dataset: Kaggle dogs-vs-cats. Contains 25,000 pictures of cats and dogs, half of them.

Development brief: Use data augmentation to improve neural network performance to increase accuracy, following this method to cope with small data sets. Data processing: convert JPEG to RGB pixel grid and convert to float tensor, and finally pack samples into batches.

Model :

Accuracy and loss charts before data augmentation and the accuracy of it :

Data augmentation :

Accuracy and loss charts after data augmentation and the accuracy of it :

Detail :

https://colab.research.google.com/drive/1WsXGadFcA8e2Afs6Sxel9kiriAzFtyRQ?usp=share_link