This is to create a repos for CNN image classifications model feasbility investigation.

1st is to build a CNN model to predict cat or dog and data augmentation is used to improved the model. Due to size limit, unable to upload train and test files Please refer to Kaggle for data source: https://www.kaggle.com/c/dogs-vs-cats

2nd one is build a CNN model to predict fashion labels using Fashion MNIST dataset. Due to size limit, unable to upload train and test files Please refer to Kaggle for data source: https://www.kaggle.com/zalando-research/fashionmnist

3rd one is to build a LE-Net model to predict traffic sign images (43 classes). The model is implemented as referred in original publication with model evaluation and model prediction.

4th one is to fine-tune ResNet and InceptionV3 models for covid and Pneumonia detections (4 classes in total). The model is created based on transfer learning with the last 10 layer trainable. Grid search is used to optimized the training parameters, including dropout rate, batch size, epoch. Dataset for training is too large to share, but happy to share on demand.

Production and deployment level codes are to be opened soon. Stay tuned.