Building Artificial Neural Networks for finding the nuclei in divergent images to advance medical discovery.
Task 1: Show understanding of the specific problem, data and its characteristics via e.g. plots and preliminary analysis.
Task 2: Implement 3 different feature engineering and/or segmentation approaches such as: {HoG (His- togram of Gradients) features, Data augmentation, Bag of visual words representations, SIFT features or free-to-use variants, Watershed segmentation}.
Task 3: Implement a mask prediction/segmentation technique that does not involve the use of Artificial Neural Networks. Use own technique to predict/segment masks in the competition.
Task 4: Implement a Multi-Layer Perceptron (Classifier) using raw pixel values (or simple functions of them) as inputs to classify pixels and predict masks. Tune the model parameters and submit the best model predictions to the competition.
Task 5: Implement Multi-Layer Perceptrons (Classifiers) using features derived from 3 feature engi-neering approaches as inputs and proceed as above. Tune model parameters and submit the best model predictions to the competition.
Task 6: Implement a Convolutional Neural Network (e.g. see this simple CNN) using raw pixel values (or simple functions of them) as inputs. Tune the model and submit the best model predictions to the competition.