Deep learning model for Identification of Dog Breed using Tensor Flow Hub

This is a beginner level project on TensorFlow and TensorFlow hub to understand the concepts and workflow of DeepLearning Models using Unstructured data.

• Getting our data ready
  • Import the data (Turning into Tensors)
  • Create Train set and Validation Set
  • Preprocess the images -> Turn them into Tensors
  • Turning into Batches
• Building a Deep Learning Model
• Training a model on few sample
• Evaluation and Making prediction on few samples
• Visualize and Tests on few samples
• Saving and Loading the Model (trained on few samples)
• Training the model in full data
• Testing and Making predictions on trained model with full data

• Google Colab, GPU usage
• Convert Images into Tensors, covered in Preprocessing of Images
• Batch & Unbatch data
• Data Visualization
• Transfer Learning
• Keras & its layers and executions
• Tensorboard Callbacks
• Early Stopping
• Epochs
• argmax(), argsort()

• Numpy
• Pandas
• Matplotlib
• TensorFlow
• TensorFlow Hub
• TensorBoard Magic Function (for visualizing performance)
• Image Classfication Model (mobilenet_v2) for Transfer Learning

Download from Kaggle

Evaluation will be done on Multi Class Log Loss between the predicted probability and the observed target.

More Details...

• Create a confusion matrix with our models predictions and true labels