ThunderNet Towards Real-time Generic Object Detection

Real-timegenericobjectdetectiononmobileplatformsisacrucialbutchallengingcomputervisiontask.However,previousCNN-baseddetectorssufferfromenormouscomputationalcost,whichhindersthemfromreal-timeinferenceincomputation-constrainedscenarios.Inthispaper,weinvestigatetheeffectivenessoftwo-stagedetect