Reconstructing three-dimensional (3D) morphology of neurons is essential to understanding brain structures and functions. Over the past decades, a number of neuron tracing tools including manual, semiautomatic, and fully automatic approaches were developed to extract and analyze 3D neuronal structures. Nevertheless, most of them were developed with conventional techniques and yield limited performance given the complicated nature of neuronal structures. Recently, Deep Learning has been shown to outperform traditional methods in a wide range of image analysis and computer vison applications. Here we developed a new open source tracing toolbox, DeepNeuron, which utilizes Deep Learning networks to trace neurons in light microscopy images. DeepNeuron provides a family of modules to solve basic yet strenuous problems in neuron tracing, including detecting neurites under different image conditions, locally connecting continuous neuronal signals, pruning and refining automatic tracing results, quantitatively evaluating manual reconstruction quality, and real-time classification of dendrites and axons. We have tested DeepNeuron using challenging light microscopy images including bright-field and confocal images of human and whole mouse brain, on which DeepNeuron demonstrates very good robustness and accuracy in neuron tracing.