In order to better apply the target detection technology to the actual production environment of domestic garbage detection, this paper proposes a lightweight domestic garbage detection method based on improved FPN. This method takes the one-stage target detection method RetinaNet as the baseline, firstly it replaces the backbone network from Resnet50 to MobilenetV1, to improve the timeliness of feature extraction in detection. Then, based on the original top-down fusion feature pyramid method, a bidirectional fusion feature pyramid method called garfpn is proposed to improve the detection accuracy. The experiments show that our method can detect domestic garbage accurately and quickly, and has good robustness and practicability.