SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation

2 Nov 2015Vijay BadrinarayananAlex KendallRoberto Cipolla

We present a novel and practical deep fully convolutional neural network architecture for semantic pixel-wise segmentation termed SegNet. This core trainable segmentation engine consists of an encoder network, a corresponding decoder network followed by a pixel-wise classification layer... (read more)

PDF Abstract

Code


Evaluation Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Semantic Segmentation ADE20K SegNet Validation mIoU 21.64 # 11
Lesion Segmentation Anatomical Tracings of Lesions After Stroke (ATLAS) SegNet Dice 0.2767 # 8
Lesion Segmentation Anatomical Tracings of Lesions After Stroke (ATLAS) SegNet IoU 0.1911 # 7
Lesion Segmentation Anatomical Tracings of Lesions After Stroke (ATLAS) SegNet Precision 0.3938 # 8
Lesion Segmentation Anatomical Tracings of Lesions After Stroke (ATLAS) SegNet Recall 0.2532 # 8
Real-Time Semantic Segmentation CamVid SegNet mIoU 46.4% # 6
Real-Time Semantic Segmentation CamVid SegNet Time (ms) 217 # 4
Real-Time Semantic Segmentation CamVid SegNet Frame (fps) 4.6 # 4
Semantic Segmentation CamVid SegNet Mean IoU 46.4% # 8
Semantic Segmentation Cityscapes test SegNet Mean IoU (class) 57.0% # 49
Real-Time Semantic Segmentation Cityscapes test SegNet mIoU 57.0% # 14
Real-Time Semantic Segmentation Cityscapes test SegNet Time (ms) 60 # 6
Real-Time Semantic Segmentation Cityscapes test SegNet Frame (fps) 16.7 # 8
Scene Segmentation SUN-RGBD SegNet Mean IoU 31.84 # 2