Figure 5
From: Machine Learning-Based Analysis of Sperm Videos and Participant Data for Male Fertility Prediction
The different machine learning-based algorithms (classical and deep learning) used to predict semen quality in terms of progressive, non-progressive, and immotile spermatozoon. The stippled line represents the threshold for the results to be considered significant compared to the ZeroR baseline. The y-axis does not start at 0 to better highlight the differences. For the methods which used dense optical flow, stride values, how many frames are skipped when comparing two frames, are presented with a 1 or 10 indicating the number of skipped frames. Dense Optical Flow (1) and Channel-wise Greyscale are the best-performing ones but, several of our proposed methods are below the significance threshold.
