Collaborative Intelligence: Accelerating Deep Neural Network Inference via Device-Edge Synergy

Table 3

Inference latency of VGGNET on four hardware architectures under different strategies.

Inference latency on	HW1 (ms)	HW2 (ms)	HW3 (ms)	HW4 (ms)
VGGNET (baseline)	69.0641	69.1523	64.2927	16.2941
VGGNET (status)	24.8501	25.1915	19.9543	19.9781
VGGNET (cogent)	7.7703	12.1214	11.4689	3.7743