important.md

This states important concepts in my notebooks

Siamese Network

This siamese network outputs two vectors and compares their euclidian distance. By this method, I compare the similiarity and dissimilarity between pairs of images of fashion items. The Fashion Mnist Dataset is used here.

Confusion Matrix

The following notebooks contain code with confusion matrices. This is my preferred evaluation technique for classification problems.

• Code 1
• Code 2

Image Models with helpful image visualization functions

• Pairs of Images and their euclidean distance
• Image, Correct Label, Predicted Label
• GIF Mnist. Red label for wrong label
• Image, Label and Accuracy of prediction
• Coloured Image with predictions
• Upload file for prediction

My Model Implementations

• ResNet
• ResNet50
• VGG Network
• Siamese Network
• Inception Resnet V2
  • Object detection
• Inception V3
• MobileNetV2
  • Object detection
• Encoder - Decoder
  • VGG 16 - FCN8
• Retina Net (Loaded from model checkpoints)
• U-Net
• Mask R-CNN

Datasets

• Fashion Mnist

  • Siamese Network
  • Custom Training Loop

• Mnist

  • ResNet
  • Mirrored Distribution Strategy
  • Object Localization

• Wine Quality

• Cat vs Dogs

  • Custom VGG model
  • Inception V3 - transfer learning

• Breast Cancer

• Horse or Human

• Flowers

• Oxford flowers 2

• CIFAR - 10

• Caltech birds