Cat and Dog CNN with ResNet50

DataSet From Kaggle - Cat and Dog

Note: dataset is too large to be included directly. Please download it yourself from the provided Kaggle link.

(also provided Traditional Chinese version document README-CH.md.)

Using ResNet50 as a feature extractor and adding additional neural network layers, the model classifies images of cats and dogs, with the final output consisting of 2 neurons representing the cat and dog classes.

Overview

• Language: Python v3.10.12
• Package: Tensorflow
• Model: CNN(ResNet50)

Model Architecture

The model uses Cross Entropy as the loss function, Adam optimizer with a learning rate of 0.0001, and applies data augmentation to reduce overfitting by generating variations of the training images.

              OPERATION        DATA DIMENSIONS   WEIGHTS(N)   WEIGHTS(%)

              Input   #####       3  224  224
         InputLayer     |      ----------------          0         0.0%
                      #####       3  224  224
      ResNet50 (Base)  \|/     ----------------    2359808         1.7%
               -      #####     512  224  224
       MaxPooling2D   Y max    ----------------          0         0.0%
                      #####     512  112  112
      Convolution2D    \|/     ----------------     147584         0.1%
               relu   #####     128  112  112
       MaxPooling2D   Y max    ----------------          0         0.0%
                      #####     128   56   56
           Flatten    |||||    ----------------          0         0.0%
                      #####         50176
              Dense   XXXXX    ----------------    1605696        74.3%
               relu   #####          32
           Dropout    |||||    ----------------          0         0.0%
                      #####          32
              Dense   XXXXX    ----------------         64         2.8%
               relu   #####           2
              Dense   XXXXX    ----------------         64         2.8%
            softmax   #####           2

Conclusion

Loss

Accuracy

Confusion Matrix - Accuracy Rate 97.53%