The invention discloses a vision-based commodity identification detection method and device, and relates to the field of article identification, and the method comprises the steps: collecting an imageof a sample commodity; according to commodity information of a preset sample commodity, marking the sample commodity in the image; carrying out deep learning on the labeled image through a CNN neuralnetwork to obtain a commodity feature model of the sample commodity; sending the commodity feature model to a preset commodity identification terminal; and the commodity recognition terminal collectsan image of a target commodity and carries out image recognition on the target commodity according to the commodity feature model to obtain commodity information corresponding to the target commodity. According to the invention, the problem of low recognition efficiency caused by the fact that the training process and the recognition process of the commodity recognition detection method in the related technology are both on one terminal device is solved.