Fig. 6From: WorMachine: machine learning-based phenotypic analysis tool for wormsWormNet architecture. The network receives an input image of size 64 × 128 pixels and outputs a classification into worm or non-worm (noise) categories. The complete architecture is detailed in the figure and can also be viewed within the WorMachine’s open-source code. WormNet was trained on 11,820 mask images of worms and non-worms, randomly split into 85% training and 15% test set. It classified with 99.2% accuracy on the training set and 93.4% on the held-out test setBack to article page