When processing the detection of boats on aerial images by neural networks, we have always been concernedabout the execution time of these networks in the equipment on board the Unmanned AerialVehicle (UAV). The UAV, throughout its mission, will capture images that must be processed in realtime. For this, an execution time optimized network is essential. This article proposes a method forsearching for optimized detection networks, for a given dataset, using an evolutionary algorithm. Theresearch uses mutations as a mechanism of evolution that affect the structure of the network and thehyper-parameters of its layers. Its fitness function allows the choice of architectures that are not verygreedy in terms of operations favoring small networks whose advantages are to be fast and quick totrain and thus to accelerate the search algorithm. Thanks to this method we were able to obtain detectionnetworks with performance similar to the initial network (parent) but with much fewer operationsallowing considerable gains in terms of execution times in an environment embedded on a drone.
When processing the detection of boats on aerial images by neural networks, we have always been concerned about the execution time of these networks in the equipment on board the Unmanned AerialVehicle (UAV). The UAV, throughout its mission, will capture images that must be processed in realtime. For this, an execution time optimized network is essential. This article proposes a method for searching for optimized detection networks, for a given dataset, using an evolutionary algorithm. The search uses mutations as a mechanism of evolution that affect the structure of the network and the hyper-parameters of its layers. Its fitness function allows the choice of architectures that are not very greedy in terms of operations favoring small networks whose advantages are to be fast and quick to train and thus to accelerate the search algorithm. Thanks to this method we were able to obtain detection networks with performance similar to the initial network (parent) but with much fewer operations allowing considerable gains in terms of execution times in an environment embedded on a drone.
When processing the detection of boats in aerial images by neural networks, we have always been concerned about the execution time of these networks in the equipment on board the Unmanned Aerial Vehicle (UAV). Throughout its mission, the UAV will capture images that must be processed in real time. For this purpose, a network optimized for execution time is essential. This article proposes an enhanced Network Architecture Search (NAS) method for searching for time-optimized detection networks, for a given dataset, using an evolutionary algorithm. The search uses mutations as a mechanism of evolution that affect the structure of the network and the hyper-parameters of its layers. Its original fitness function allows the choice of architectures that are not very greedy in terms of operations, specifically favouring small networks whose advantages are to be fast and quick to train, thus accelerating the search algorithm. Using this method, we were able to obtain detection networks with an improved mean Average Precision (mAP) compared to the initial network (parent) but with much fewer FLoating-point OPerations (Flops): 68% of operations reduction. This induces considerable gain in terms of execution time with 50 Frames Processed per Second (FPS) in an embedded environment on a drone.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.