The invention discloses a power artificial-intelligence visual-analysis system based on multi-core heterogeneous parallel computing. The system is characterized by including a lightweight neural network as a core, and including a multi-core heterogeneous parallel computing module and a business application module. Data are accessed and transmitted between the business application module and the multi-core heterogeneous parallel computing module through a web service interface. The multi-core heterogeneous parallel computing module includes a GPU (Graphic Processing Unit) computing node, a CPU(Central Processing Unit) storage management node and a CPU computing node. All the nodes are connected through a switch. The business application module includes an image management module, an imagelabeling module, a model training module and an algorithm application module. The system is based on a multi-core heterogeneous parallel computing framework, can highly efficiently realize online or offline image data training to form a lightweight fast image classification model to realize power intranet/extranet image business application, can exert values of image data in a maximized manner, and has better application prospects.