Table 3. An ablation test of Selfee training process on mice datasets.

Model	Single frame + KNN		Live-frame + KNN		Single frame + LGBM		Live-frame + LGBM
Evaluation	Mean F1 score	Mean AP	Mean F1 score	Mean AP	Mean F1 score	Mean AP	Mean F1 score	Mean AP
Replication 1	0.554	0.498	0.590	0.530	0.645	0.671	0.674	0.691
Replication 2	0.574	0.508	0.599	0.549	0.653	0.663	0.663	0.699
Replication 3	0.566	0.514	0.601	0.539	0.652	0.692	0.663	0.700
Mean	0.565	0.507	0.597	0.539	0.650	0.675	0.667 *	0.697 *
Best	0.574	0.514	0.601	0.549	0.653	0.692	0.674 *	0.700 *

^*Best results of different training setups under each evaluation metric are indicated in bold values.