RnDProjectsDeebul / ManojKolpeThesis

Mozilla Public License 2.0
2 stars 1 forks source link

Extrinsic parameters of the cityscape dataset #10

Open Manojkl opened 2 years ago

Manojkl commented 2 years ago

While looking at the intrinsic and extrinsic parameters data, the extrinsic parameter did not change; it was constant for all the images. Then I looked into the datasheet they have provided, so the extrinsic parameter that they have provided is concerning the vehicle coordinate system. Attaching the image below for reference. But we need the extrinsic parameter concerning the world coordinate system, which we don't have. But we have the longitude and latitude information of the car. image

So I think we can train the U-net model with the cityscape data without temporal fusion and then collect our data or look for other data with an extrinsic parameter for fusion. I can train the segmentation network with the cityscape data but no temporal fusion since I don't have the extrinsic parameter of the images concerning the world coordinate system. We can take the weights of the encoder-decoder for training with our data.