Table 4.

SSL (co-training) results on vehicle (V) and pedestrian (P) detection, reporting mAP. From a training set ${\mathcal{X}}^{tr}\in \{{\mathcal{K}}^{tr},{\mathcal{W}}^{tr}\}$, we preserve the labeling information for a randomly chosen p% of its images, while it is ignored for the rest. We report results for p = 100 (all labels are used), p = 5 and p = 10. If ${\mathcal{X}}^{tt}={\mathcal{K}}^{tt}$, then ${\mathcal{X}}^{tr}={\mathcal{K}}^{tr}$; analogously, when ${\mathcal{X}}^{tt}={\mathcal{W}}^{tt}$, then ${\mathcal{X}}^{tr}={\mathcal{W}}^{tr}$, i.e., there is no domain shift in these experiments. Co-T (RGB) and Co-T (RGB/D) stand for single and multi modal co-training, respectively. UP and LB stand for upper bound and lower bound, respectively. Bold results indicate best performing within the block, where blocks are delimited by horizontal lines. Second best is underlined, but if the difference with the best is below 0.5 points, we use bold too. $\mathsf{\Delta }\left\{{\varphi }_{F_1}\mathrm{vs}.{\varphi }_{F_2}\right\}$ stands for mAP of ${\varphi }_{F_1}$ minus mAP of ${\varphi }_{F_2}$.

Training Set	${\mathcal{X}}^{\mathit{tt}}={\mathcal{K}}^{\mathit{tt}}$			${\mathcal{X}}^{\mathit{tt}}={\mathcal{W}}^{\mathit{tt}}$
	V	P	V&P	V	P	V&P
100% Labeled (RGB)/UB	83.43	67.77	75.60	61.71	57.74	59.73
100% Labeled (D)/UB	80.80	53.43	67.12	55.14	37.67	46.41
5% Labeled (RGB)/LB	65.20	46.08	55.64	51.69	41.92	46.81
5% Labeled (D)/LB	64.45	26.70	45.58	45.21	29.98	36.70
5% Labeled + Co-T (RGB)	74.26	55.41	64.84	54.00	56.34	55.17
5% Labeled + Co-T (RGB/D)	78.64	57.40	68.02	58.42	56.98	57.70
10% Labeled (RGB)/LB	72.31	45.51	58.91	49.53	49.83	49.68
10% Labeled (D)/LB	69.54	46.31	57.93	47.93	33.98	40.96
10% Labeled + Co-T (RGB)	78.63	60.99	69.81	56.15	60.20	58.18
10% Labeled + Co-T (RGB/D)	79.68	60.55	70.12	59.54	57.17	58.36
$\mathsf{\Delta }\left\{(5\%\mathrm{L}.+\mathrm{C}\mathrm{o}-\mathrm{T}(\mathrm{R}\mathrm{G}\mathrm{B}/d\left)\right)\mathrm{v}\mathrm{s}.(5\%\mathrm{L}.(\mathrm{R}\mathrm{G}\mathrm{B})/\mathrm{L}\mathrm{B})\right\}$	+13.44	+11.32	+12.38	+6.73	+15.06	+10.89
$\mathsf{\Delta }\left\{(5\%\mathrm{L}.+\mathrm{C}\mathrm{o}-\mathrm{T}(\mathrm{R}\mathrm{G}\mathrm{B}/d\left)\right)\mathrm{v}\mathrm{s}.(100\%\mathrm{L}.(\mathrm{R}\mathrm{G}\mathrm{B})/\mathrm{U}\mathrm{B})\right\}$	−4.79	−10.37	−7.58	−3.29	−0.76	−2.03
$\mathsf{\Delta }\left\{(10\%\mathrm{L}.+\mathrm{C}\mathrm{o}-\mathrm{T}(\mathrm{R}\mathrm{G}\mathrm{B}/d\left)\right)\mathrm{v}\mathrm{s}.(10\%\mathrm{L}.(\mathrm{R}\mathrm{G}\mathrm{B})/\mathrm{L}\mathrm{B})\right\}$	+7.37	+15.04	+11.21	+10.01	+7.34	+8.68
$\mathsf{\Delta }\left\{(10\%\mathrm{L}.+\mathrm{C}\mathrm{o}-\mathrm{T}(\mathrm{R}\mathrm{G}\mathrm{B}/d\left)\right)\mathrm{v}\mathrm{s}.(100\%\mathrm{L}.(\mathrm{R}\mathrm{G}\mathrm{B})/\mathrm{U}\mathrm{B})\right\}$	−3.75	−7.22	−5.48	−2.17	−0.57	−1.37