无需训练的运动统计特征在动物行为基准上超越自监督编码器
Training-free motion statistics match self-supervised behavior encoders
这篇论文做了个挺扎心的实验:一套不用训练的 PCA 统计特征,在 CalMS21、MABe22 这些基准上把 BAMS 和 hBehaveMAE 都比下去了。做动物行为分析或时序表征的研究者应该看看,至少以后跑实验前先报一下这个基线。
一篇预注册协议下的对比研究显示,用逐帧运动测量按固定时间窗口汇总并经 PCA 降维的训练无关表征,在多个动物行为基准上超过了已发布的自监督编码器。
在 CalMS21 上它达到 85.8 macro F1,高于最佳编码器的 82.
4;在 MABe22 小鼠三连组上得 83.
7 对 hBehaveMAE 的 80.3(差距 +3.
4,95% 区间 +2.1 到 +4.
7);在 MABe22 果蝇群体 50 个任务上得 51.
6 平均 F1,超过 BAMS 的 48.2。
对比采用等大小冻结表征加单一共享分类器,并用三个随机种子重训编码器验证结论稳健。机制分析发现短时间窗口贡献了大部分结果:以每个身体自身坐标系表示的原始关键点在相同窗口下汇总,已在两个小鼠数据集和 Shot7M2 上追平或超越编码器,只有果蝇任务还需额外测量补充 5.
3 分。研究者的结论是:在这些基准上,现有行为编码器尚未超过它们所学习的手工特征,论文呼吁评测时同时报告这类基线。
Training-free motion statistics match self-supervised behavior encoders
Self-supervised encoders for animal behavior are often trained on, or fed, hand-built movement features. We ask whether they beat those ingredients when compared fairly. A training-free representation, made of generic per-frame measurements summarized over fixed time windows and reduced with PCA, is compared with retrained and released encoders (BAMS, hBehaveMAE) under a preregistered protocol: frozen representations of equal size and one shared classifier. On CalMS21 it reaches 85.8 macro F1 against 82.4 for the best encoder, ties with four labelled videos, and leads when the annotator changes or new behaviors must be learned from a few videos; the reading holds for encoders retrained with three seeds. On the MABe22 mouse triplets it scores 83.7 against 80.3 for the released hBehaveMAE (gap +3.4, 95% interval +2.1 to +4.7). On a second species, the MABe22 fly groups, the same recipe scores 51.6 mean F1 over 50 tasks against 48.2 for BAMS, the best published method. On Shot7M2, the synthetic basketball benchmark built by hBehaveMAE's authors, it scores 68.8 against 64.7 for hBehaveMAE, the best method in that paper, though hBehaveMAE remains better on the shortest movements. Short time windows carry most of the result. Raw keypoints in each body's own frame, summarized over the same windows, already match or beat the encoders on both mouse datasets and on Shot7M2; on the fly groups they fall short, and the measurements add 5.3 points. On these benchmarks, current behavior encoders have yet to beat the hand-built features they learn from, and such baselines should be reported alongside them.
本条情报归属 AITOP「论文」核心追踪栏目。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:BAMS、hBehaveMAE、CalMS21、MABe22,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明:本文原始线索采编自「arXiv cs.LG」(@Ziwen Chen),经 AITOP 多维评分管线结构化提炼。