How to understand SynWoodScape extrinsic translation used in CARLA?
dengping023 opened this issue · 1 comments
SynWoodScape and WoodScape mount the camera in a similar way, they have the almost same extrinsic-quaternion，but the translations are different which is confusing.
SynWoodScape-FV,RV,MVL,MVR:
"translation used in CARLA (CARLA reference)": [1.92, 0.0, 0.9]
"translation used in CARLA (CARLA reference)": [1.92 ,0.0, 0.9]
"translation used in CARLA (CARLA reference)": [0.8, 1.0, -0.9 ]
"translation used in CARLA (CARLA reference)": [0.8, 1.0, 0.9 ]
WoodScape-FV,RV,MVL,MVR:
"translation": [3.7484, 0.0, 0.6577999999999999 ]
"translation": [-1.0678, 0.0545, 0.9303899999999999]
"translation": [1.9445999999999999, 0.9572999999999999, 0.95874 ]
"translation": [1.9445999999999999, -0.9572999999999999, 0.97305 ]
I try to make 4 fishimg to a bev img, its ok in WoodScape dataset，but can not make it throught in SynWoodScape
Thank you for your question. The sensors are attached to the ego vehicle following the reference in the image bellow.
And indeed there is a problem for camera MVL and MVR here is the correct positions in meters:
| Camera | x | y | z |
|---|---|---|---|
| FV | 1.92 | 0.0 | 0.9 |
| RV | -1.92 | 0.0 | 0.9 |
| MVR | 0.8 | 0.9 | 1.0 |
| MVL | 0.8 | -0.9 | 1.0 |
