Image-data-analysis

In this project we used Deep learning algorithms to achive pathologist-lavel performance for detecting the Gleason grade of prostate tissue samples. Finally, we obtained the results using the efficientnet by applying following techniques: We did tile preparation on the dataset which is based on the public kernel. For augmentation, we add cutouts, rotations augmentation for better generalization and mix-up augmentation for training better generalization. Here we trained the 5-folds model. Hence, we obtained the following information from the datasets using the convolutional neural network architecture of efficientnet: image id, data provider, isup grade, gleason score, kfold etc. So, now we can figure out the benign or cancer, and grade 1 to grade 5. Please see the attached .txt file for details information on the results.