小模型达99.7%准确率，突破手写天城文识别性能极限。
Barnamala: Parameter-Efficient Handwritten Devanagari Recognition at Benchmark Saturation
- 仅110万参数的轻量卷积网络，实现46类天城文识别
- 所有模型均逼近11错误的固有上限，无法显著超越
- 零样本识别数字达76.6%，抗干扰能力远超大模型
我们构建了一个仅110万参数的紧凑卷积网络，用于46类DHCD天城文识别，达到99.73%的准确率，是此前最优模型（1732万参数）的15.6倍更小。经验证，当前已达到性能饱和：所有测试模型（含大型教师集成）均趋近相同的11错误固有下限，无配置在精确麦克奈马尔检验与威尔逊置信区间下具有统计显著优势。即使不使用知识蒸馏，该学生模型仍与最接近的大模型基线持平（麦氏检验p=0.345）。在非DHCD数据集上，零样本识别CMATERdb数字达76.6%，微调后可达97.8%；抗干扰能力亦显著优于大模型（平均扰动准确率75.7% vs. 38.7%）。所有代码与结果见https://github.com/Ampixa/barnamala。
原文摘要 · Abstract (English)
We built a compact convolutional network (1.11 M parameters) for 46-class DHCD Devanagari recognition and reached 99.73%, the highest reported at 15.6x smaller than prior state-of-the-art. We have effectively reached the saturation point: every model tested, large teacher ensembles included, hits the same 11-error intrinsic floor. No configuration achieves a statistically clear win under exact McNemar tests with Wilson confidence intervals. Even without knowledge distillation, our student matches the nearest large-model baseline (17.32 M parameters; McNemar $p = 0.345$). Outside of DHCD, zero-shot on CMATERdb digits gives 76.6% and fine-tuning reaches 97.8%; corruption robustness is also far better than large baselines (mean corruption accuracy 75.7% vs. 38.7%). All artifacts are at https://github.com/Ampixa/barnamala.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。