Fig. 5: Deep learning-assisted single-atom detection of copper ions.

a Negative and positive FSV signals. The curves shown here are a subset of FSV voltammograms without averaging by oscilloscope and background subtraction. The theoretical number of Cu²⁺ contained in the sample is 1.2. b The framework of our image classifier FSVNet based on DCNN. FSV voltammogram images are captured by the home-built electrochemical system and inputted to the computer for deep learning, including feature extraction and classification predictions. c Accuracy and d loss rates of FSVNet during the learning process. e AUC diagram of FSVNet model on testing dataset. The receiver operating characteristic (ROC) curve illustrates an area under the curve (AUC) of 1.00. Source data are provided as a Source Data file.