Table 2.

Performance evaluation metrics of the MobileNetv3-large model on the test set

Category	Precision	Sensitivity	Specificity	f1-score	Accuracy	AP	AUC	MCC	Cohen’s kappa
Angiodysplasia	0.9279	0.9115	0.9911	0.9196	0.9115	0.9664	0.9885	0.9097	0.9097
Bleeding	0.9808	0.9027	0.9978	0.9401	0.9027	0.9851	0.9977	0.9339	0.9329
Erosion	0.646	0.8902	0.957	0.7487	0.8902	0.8396	0.9791	0.7341	0.7227
Erythema	0.8431	0.7679	0.9916	0.8037	0.7679	0.8676	0.9796	0.7938	0.7928
Foreign Body	0.9717	1	0.9967	0.9856	1	0.9998	1	0.9841	0.984
Lymph Follicle	0.859	0.8072	0.9882	0.8323	0.8072	0.9224	0.99	0.8183	0.8178
Lymphangiectasia	0.8615	1	0.9906	0.9256	1	0.9977	0.9999	0.9238	0.9209
Normal mucosa	0.7731	0.9293	0.9705	0.844	0.9293	0.9483	0.9895	0.8298	0.8254
Polyp	0.8788	0.7436	0.9866	0.8056	0.7436	0.8584	0.9766	0.786	0.7825
SMT	0.9667	0.725	0.999	0.8286	0.725	0.9511	0.997	0.8317	0.8226
Stenosis	0.925	0.7115	0.9969	0.8043	0.7115	0.9202	0.9912	0.8027	0.7952
Vein	0.9381	0.9192	0.9934	0.9286	0.9192	0.9566	0.9923	0.921	0.9209
macro avg	0.881	0.859	0.9883	0.8639	0.859	0.9344	0.9901	0.8557	0.8523
weighted avg	0.883	0.8717	0.9877	0.8724	0.8717	0.9361	0.9897	0.8631	0.8602

Note The macro average metric treats all categories equally, ensuring that even categories with smaller sample sizes contribute equally to the model’s performance evaluation. In contrast, the weighted average metric considers the sample size of each category, assigning greater weight to categories with larger sample sizes