Object separation method and system based on lightweight detection

The invention discloses an object separation method based on lightweight detection. The method comprises the following steps: acquiring an object image; inputting the image into a lightweight backbone network, extracting features, and obtaining a feature map with the same scale; obtaining a multi-scale feature map from the equal-scale feature maps through a deep and shallow feature fusion model; performing multi-scale feature fusion by using a multi-scale fusion-based double-attention algorithm based on the multi-scale feature map to obtain a plurality of different deep-layer small-scale detection branches and a plurality of different shallow-layer large-scale detection branches; constructing a prediction branch for each detection branch to detect an object under the scale, and merging and post-processing output feature vectors of the plurality of prediction branches to obtain a detection result of the object; and performing object separation according to a detection result. The invention further discloses a