kristin-zlchen/multimodalseg

Multi-modal semantic segmentation

PyTorch models

Dataset

PASCAL VOC

model	acc	acc_cls	mean_iu	notes
FCN32s	90.17%	75.56%	61.81%	lr=1.0e-10 reduction='sum'
FCN32s(original)	-	-	63.6%
FCN8sAtOnce	90.27%	74.95%	62.13%	lr=1.0e-10 reduction='sum'
FCN8sAtOnce(original)	-	-	65.4%
DeepLab-LargeFov	93.71%	72.21%	61.32%	pad images to 513x513 for evaluation
DeepLab-LargeFov	90.90%	73.89%	62.09%	use original resolution for evaluation
DeepLab-LargeFov(original)	-	-	62.25%
DeepLab-ASPP	93.10	80.13%	61.07%
DeepLab-ASPP(original)	-	-	68.96%

CamVid

model	acc	acc_cls	mean_iu	notes
SegNet(Maxunpooling, vgg16-based)	86.71%	66.39%	54.09%	lr=0.01
SegNet(Maxunpooling, vg16_bn-based)	87.84%	70.75%	57.68%	lr=0.01
SegNet(Bilinear interpolation)	85.86%	71.95%	56.22%	lr=0.01
SegNet(original)	88.6%	65.9%	50.2%
UNet	84.38%	62.80%	49.83%	lr=0.01