发表机构
Foundation for Computational Science (FOCUS)（计算科学基金会（FOCUS））
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究通过14种输入分辨率、六种架构及边缘设备实测，发现目标精度决定最优分辨率，更换模型比提高分辨率更有效，并揭示了预处理路径、精度引擎及FP16数值范围对延迟和精度的重要影响。
AI 中文摘要
风电场鸟类撞击缓解需要识别仅占几十像素的远处鸟类，因此分类器的输入分辨率N是一个设计变量，而非固定规格。我们通过一个因子设计对其进行研究，涵盖14种边长N（16至224）、六种架构、两种训练与评估机制以及30个随机种子——共2,520个检查点和5,040次评估——并在NVIDIA Jetson Orin Nano上测量延迟。得到四项结果。(1) 所选N取决于目标：ResNet50在N=112时于验证集上达到0.90，估计耗时1.85毫秒（测试集上为0.8980）；DINOv2-L在N=144时达到0.95，耗时12.70毫秒；更换模型比提高N能带来更多精度提升（在N=112时，+5.93对比+2.33个百分点）。(2) 降低N的收益取决于假定的预处理路径：当每个个体从其自身文件解码时，N从224降至80节省13.5%的时间；但当检测器一次性解码4K帧时，节省46.2%的时间；帕累托最优集从20个配置增加到23个。(3) 精度必须在部署引擎上测量：半精度单独使ViT-S/16在N>=96时损失4至7个百分点，而CNN的损失保持在0.1个百分点以内；在验证时保持选择的情况下，176个目标中有26个的选择不同。一个损坏的FP16引擎可能比正确的引擎运行得更快，这从延迟上无法察觉；允许14个ViT-S/16 FP32配置将推荐结果移至0.931-0.938区间内，并满足10毫秒预算。(4) ViT-L规模的模型适合此设备，但激活值超出FP16范围；在Transformer块边界处分割计算图，将FP32限制在受影响的段内，使部署的DINOv2-L链比单引擎构建快1.85倍。我们还量化了机制差异符号如何随种子数量而稳定；一种敏感性分割移除了某些形式的组共享，在选择边界处保留了所有14个非平凡符号。
英文摘要
Bird-strike mitigation at wind farms requires identifying distant birds that span only tens of pixels, so the classifier's input resolution N is a design variable, not a fixed specification. We study it with a factorial design over 14 side lengths N (16 to 224), six architectures, two training and evaluation regimes and 30 random seeds -- 2,520 checkpoints and 5,040 evaluations -- plus latency measured on an NVIDIA Jetson Orin Nano. Four results. (1) The selected N depends on the target: 0.90 is met on validation by ResNet50 at N=112 in an estimated 1.85 ms (0.8980 on test) and 0.95 by DINOv2-L at N=144 in 12.70 ms; changing the model buys more accuracy than raising N (+5.93 versus +2.33 points at N=112). (2) The benefit of lowering N depends on the assumed preprocessing path: N=224 -> 80 saves 13.5% when each individual is decoded from its own file but 46.2% when the detector decodes the 4K frame once; the Pareto set grows from 20 to 23 configurations. (3) Accuracy must be measured on the deployed engine: half precision costs ViT-S/16 alone 4 to 7 points at N>=96 while the CNNs stay within 0.1 points, and with selection held at validation the choice differs at 26 of 176 targets. A broken FP16 engine can run faster than a correct one, undetectable from latency; admitting 14 ViT-S/16 FP32 configurations moves the recommendation over the 0.931-0.938 band and under the 10 ms budget. (4) ViT-L-scale models fit this device, but activations exceed the FP16 range; splitting the graph at transformer-block boundaries confines FP32 to the affected segments, making the deployed DINOv2-L chain 1.85x faster than the single-engine build. We also quantify how the regime-difference sign stabilises with seed count; a sensitivity split removing some forms of group sharing preserves all 14 non-trivial signs at the selection boundary.
Comments25 pages, 7 figures, 7 tables. On-device measurements on NVIDIA Jetson Orin Nano (JetPack 6.2, TensorRT 10.3). Data and code: https://doi.org/10.5281/zenodo.22697995
