Viewpoint Change

EXPERIMENT 1
From lower to higher viewpoint

It tests the generalization and adaptation performances of the semantic segmentation networks when moving from a lower to an higher perspective, with a different shape of the hood. The distance between the source and target scenario is low so, as a consequence, the task is easy.

Source Scenario:

  • Environment: Town 01
  • Weather and illumination condition: Clear Sunset
  • Viewpoint: Audi TT
  • Source train set size: 10044
Source Example Source Example Source Example Source Example Source Example Source Example
Click to see some sample taken from the source scenario.

Target/Test Scenario:

  • Environment: Town 01
  • Weather and illumination condition: Clear Sunset
  • Viewpoint: Jeep Wrangler Rubicon
  • Target train set size (only with DA): 10044
  • Test size: 1674
Target Example Target Example Target Example Target Example Target Example Target Example
Click to see some sample taken from the target/test scenario.

Results

Experiment Distance Measurement Performance Evaluation
Euclidean
distance		 Cosine
distance		 Bhattacharaya
distance		 Network Code Available mIoU (%)
Source:
Town 01, Clear Sunset, Audi

Target:
Town 01, Clear Sunset, Jeep		 6,4551 1,0586 0,0426 without
domain
adaptation 		DeepLab V2[1] (soon) 62,60
DeepLab V3+ [2] (soon) 64,93
PSPNet [3] (soon) 67,32
PSANet [4] (soon) 66,88
DeepLab V2 [1]
(source=target)		 (soon) 79,13
with
domain
adaptation		 DADA [5] (soon) 66,42
ADVENT [6] (soon) 68,43
CLAN [7] (soon) 70,30
DISE [8] (soon) 73,64

Updated: