Generate dataset for visual relocalization in CARLA 0.8.2
-
Find the Two CARLA scripts, then add it to CARLA_0.8.2/PythonClient/scripts/CARLA
-
Start carla simulator in Town01 or Town 02 (default Town01) with PowerShell
.\CarlaUE4.exe -carla-server -windowed -ResX=400 -ResY=300 -benchmark -fps=10 .\CarlaUE4.exe /Game/Maps/Town02 -carla-server -windowed -ResX=400 -ResY=300 -benchmark -fps=10
-
Set up weather, playerstart, vehicles, pedestrians etc. then run RelocalizationReference.py
-
Set up weather, playerstart, vehicles, pedestrians etc. then run RelocalizationQuery.py
-
Output format
| folder name | description |
|---|---|
| e.g. W000_P100_V000_P000 | weather, playerstart, vehicles, pedestrians |
| W000_P100_V050_P200 | as above |
| W000_P100_V075_P300 | as above |
| subfolder name | description |
|---|---|
| /Depth | key frame,256*256 |
| /RGB | key frame,256*256 |
| /SemanticSegmentation | key frame,256*256 |
| Control.txt | each frame,256*256 |
| Trajectory.txt | each frame,256*256 |
-
Output format
| folder name | description |
|---|---|
| /AB | dynamic,static, width=256*2, height=256 |
| /ABC | dynamic,static,segmantic mask, width=256*3, height |
- Run test_with_mask to infer, then download to local pc from remote server
pretrained Dynamic2static model
- Output format
| folder name | description |
|---|---|
| /images | fake_B, Gx, real_A,.... |
deltete Gx, real_A,..., rename fake_B.png to xxxxxx.png
-
In pycharm run ImageRetrieval.py
-
in jupyter notebook run get_precision.ipynb (recommended)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent