Fig. 3
From: Non-resonant background removal in broadband CARS microscopy using deep-learning algorithms
Performances of the 6 models (GAN, CNN + GRU, VECTOR, SpecNet, LSTM, BI-LSTM) on a test set made of 1000 instances after training with the synthetic dataset. (a) Coefficient of determination R2. (b) Average mean square error (MSE) on the test instances. (c) The scatter plot reports the number of parameters and the mean prediction time for each model on a single instance of the test dataset.
