INF368-EXS2

Task 1

• Choose a deep neural network architecture that is known to achieve good performance for image recognition.
  • Sugguestion: YOLOv2 or v3
• You will provide in-depth information about this architecture.
• Implement it using the tensorflow.keras API.
  • cfg to h5 for any choosen yolo model.
• Verify in practice the ability of your model to recognize images on the SVHN dataset (SVHN is similar in flavor to MNIST, but for image recognition tasks - see http://ufldl.stanford.edu/house

Task 1

1. Describe architecture
2. implement model with weights
3. test model with data

Task 2

1. Datascience
2. Data generator
3. Create model

• Cut of head and twick
• Model selection on the head

4. Train on the given dataset