Fig. 9From: Intelligent fluorescence image analysis of giant unilamellar vesicles using convolutional neural networkCNN-based vesicle selection filter. a Typical vesicle images used to train the neural network. Class 1 represents vesicles that are suitable for analysis. Class 2 represents multi-lamellar vesicles. Class 3 represents vesicles that are too closely overlapped with other vesicles, and class 4 represents vesicle images that are too hazy to analyze. b, c Example results of deep learning-based vesicle selection. Class 1—cyan, Class 2—green, Class 3—blue, Class 4—magenta, vesicles at the edges—red. Invalid vesicles were identified successfully for exclusion. The images are 127.3 μm in width and heightBack to article page