The KITTI Vision Benchmark Suite

Stereo matching is a fundamental task in computer
vision, aiming to estimate depth information by
matching corresponding features between image
pairs. In conventional approaches, cost
aggregation is typically performed through
cascaded 3D convolutional layers, which leads to
substantial computational complexity and limits
practical deployment.

In this study, we leverage multi-scale features
extracted by the backbone network to guide the
cost aggregation process and propose an efficient
gated-attention-based cost aggregation module,
termed GACVE. The proposed module significantly
enhances aggregation efficiency while maintaining
competitive accuracy. Experimental results
demonstrate that only a few GACVE modules combined
with lightweight 2D cost aggregation layers are
sufficient to achieve competitive disparity
estimation performance.

Furthermore, the proposed model is evaluated on
Jetson Nano and compared with several state-of-
the-art methods, showing superior inference speed
on edge

\lr=5e-3