Town Change

EXPERIMENT 2
From a city to a bucolic country

It tests the generalization and adaptation performances of the semantic segmentation networks when changing completely the environment, moving from a city to a countryside. The distance between the source and target scenario is high so, as a consequence, the task is difficult. The task is even more challenging due to the rain condition.

Source Scenario:

  • Environment: Town 01
  • Weather and illumination condition: Hard Rain Noon
  • Viewpoint: Audi TT
  • Source train set size: 10008
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 07
  • Weather and illumination condition: Hard Rain Noon
  • Viewpoint: Audi TT
  • Target train set size (only with DA): 10008
  • Test size: 1672
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, Hard Rain Noon, Audi

Target:
Town 07, Hard Rain Noon, Audi		 6,4551 1,0586 0,0426 without
domain
adaptation 		DeepLab V2[1] (soon) 21,65
DeepLab V3+ [2] (soon) 14,27
PSPNet [3] (soon) 14,64
PSANet [4] (soon) 15,52
DeepLab V2 [1]
(source=target)		 (soon) 78,02
with
domain
adaptation		 DADA [5] (soon) 36,48
ADVENT [6] (soon) 39,30
CLAN [7] (soon) 41,18
DISE [8] (soon) 46,71

Updated: