Architecture 1. Extended efficient layer aggregation networks VovNet → CSPVoVNet The gradient analysis approach described above makes inferences faster and more accurate CSPVoVNet → ELAN By controlling the shortest longest gradient path, a deeper network can learn and converge effectively ELAN → E-ELAN E-ELAN uses expand, shuffle, merge cardinality* to achieve the ability to continuously enhance..
