Deployment of modern detection models is difficult for infraredtargetdetection used in robot vision systems due to their heavy computational burden. To alleviate this situation, a simple but efficient channel prunin...
Deployment of modern detection models is difficult for infraredtargetdetection used in robot vision systems due to their heavy computational burden. To alleviate this situation, a simple but efficient channel pruning method is proposed for model acceleration. Specifically, a soft-gated module combined with batch normalization (SGBN) is designed as a standalone layer to substitute the standard batch normalization (BN) layer during training. The conversion between SGBN and BN is easy, and the training overhead introduced is almost negligible after replacement. By controlling the sparsity of the scaling factor in SGBN, unimportant channels with small output are blocked automatically and globally, which is simultaneous with model training. Removing these redundant channels no longer requires fine-tuning, thus significantly speeding up the pruning process. Experiments of pruning different detection models on the infrared dataset show the effectiveness of our method. For example, the parameters and FLOPs of pruned CetnerNet are reduced by 72.70% and 40.20%, respectively, without accuracy loss. The inference speed on the CPU is 12.01ms faster. Extended studies on the classification task also demonstrate its great potential when transferring to other applications.
Robust ship AAW defense capability is a priority requirement that enables Naval Forces to conduct joint expeditionary force operations in littoral environments. As an aid to achieving this capability for all ship clas...
Robust ship AAW defense capability is a priority requirement that enables Naval Forces to conduct joint expeditionary force operations in littoral environments. As an aid to achieving this capability for all ship classes, the Navy has reorganized its management of ship defense. A major focus of these efforts is the development of a fully automatic, integrated combat system for non-Aegis ships which is based on coordinated detection, control, and hard kill/electronic warfare (HK/EW) engagement functions. A phased approach to attaining this fully integrated capability has been established which includes major element upgrade introduction when technology and budget permit. This paper describes the ship self-defense performance assessment methodology which has been adopted to support the review and decision process for future planning and budgeting. This process is a continuation and refinement of that used to provide data for the Office of the Secretary of Defense Fall 1991 Conventional Systems Committee Ship Self-Defense Review. The performance assessment methodology starts with definition of survivability requirements by ship class, combat system configurations, Anti-Ship Missile threat, and operational scenario. Viable self-defense system element options are then identified. The capability of these options are then characterized for input into ship level performance prediction models. Three partitions of performance prediction modeling are made: hard kill elements only, electronic warfare element only, and integrated HK/EW. The most significant accomplishment of this effort, beyond providing data to support programmatic decision, has been the creation of a truly integrated HK/EW surface ship combat system model that interleaves HK and EW timelines and allows parametric variation to evaluate options. Because of classification, only generic examples of numerical result will be presented. However, the procedures for establishing the modelling process, prioritization of
暂无评论