Training results: result_yolo26

[MateoLostanlen]

Evaluation results on test set

Metric	main	current	Δ
mAP@50	0.7556	0.7496	-0.0060
mAP@50-95	0.4802	0.4422	-0.0380
Precision	0.7886	0.7464	-0.0423
Recall	0.7361	0.6850	-0.0510

Sequential evaluation on test (optimized params: nb_frames=8, conf=0.15)

TP=114 FN=3 FP=9 TN=114

Metric	main	current	Δ
Recall (TP rate)	0.9597	0.9744	+0.0147
FPR	0.1409	0.0732	-0.0677
F1	0.9137	0.9500	+0.0363
Precision	0.8720	0.9268	+0.0548

Engine parameter optimization (top 10 on val)

nb_frames	conf	TP	FN	FP	TN	Recall	FPR	F1
8	0.15	98	5	11	113	0.9515	0.0887	0.9245
5	0.3	97	6	10	114	0.9417	0.0806	0.9238
7	0.25	97	6	10	114	0.9417	0.0806	0.9238
6	0.15	99	4	13	111	0.9612	0.1048	0.9209
7	0.15	99	4	13	111	0.9612	0.1048	0.9209
8	0.2	97	6	11	113	0.9417	0.0887	0.9194
8	0.25	96	7	10	114	0.932	0.0806	0.9187
6	0.3	95	8	9	115	0.9223	0.0726	0.9179
8	0.1	100	3	15	109	0.9709	0.121	0.9174
7	0.3	94	9	8	116	0.9126	0.0645	0.9171

(vs last committed results on main)
Branch: result_yolo26 | Test dataset: pyronear/pyro-dataset @ v2.2.0