Human Protein Atlas Competition
Images visualizing proteins in cells are commonly used for biomedical research, and these cells could hold the key for the next breakthrough in medicine. However, thanks to advances in high-throughput microscopy, these images are generated at a far greater pace than what can be manually evaluated. Therefore, the need is greater than ever for automating biomedical image analysis to accelerate the understanding of human cells and disease.
Developed model capable of classifying mixed patterns of proteins in microscope images. This is a multilabel image classification problem, where each image can belong to several classes. The class labels are as follows: 0: 'Mitochondria', 1: 'Nuclear bodies', 2: 'Nucleoli', 3: 'Golgi apparatus', 4: 'Nucleoplasm', 5: 'Nucleoli fibrillar center', 6: 'Cytosol', 7: 'Plasma membrane', 8: 'Centrosome', 9: 'Nuclear speckles'