Abstract: Faults play a crucial role in hydrocarbon migration and accumulation, making the efficient and accurate seismic fault identification essential for oil and gas exploration and development. Traditional manual interpretation methods are inefficient and highly subjective, failing to meet practical demands. The application of deep learning has significantly improved the efficiency of seismic data interpretation; however, in complex structural settings, challenges such as insufficient fault delineation still persist. To address this limitation, this study proposes the Amplitude and Ant-tracking Attribute Fusion Network (AmpAnt-Net), a dual-branch parallel encoder network that performs fault detection using both amplitude and ant-tracking attributes. The amplitude branch employs convolutional modules to extract texture features and event continuity from amplitude data, while the ant-tracking branch leverages a graph convolutional network to model the spatial connectivity and directional consistency of ant-tracking attributes. A cross-attention mechanism is introduced in the bottleneck layer to achieve effective fusion and enhancement of the two feature types. During decoding, multi-scale features are progressively integrated to reconstruct fault structures. Experimental results demonstrate that the proposed method improves IoU and F1 scores by 2.91 and 2.04 percentage points over UNet, and by 5.38 and 3.79 percentage points over UNet++, respectively, on synthetic datasets. Moreover, AmpAnt-Net exhibits superior performance in identifying small-scale faults and characterizing complex fracture zones, showing strong generalization ability and application potential.