Fig. 10From: Large-scale monocular FastSLAM2.0 acceleration on an embedded heterogeneous architectureCPU-GPU data transfer timeBack to article page