Bowl View using Fisheye Cameras from a Car

[shreshtashetty]

I am using 640x360 images from the SVS cameras of a car with this repo. I have referred to #793, but many of my doubts have not been cleared. My questions are as follows:-

1. How do I use this repository with 640x360 images instead of the standard 1920x1080? What changes do I have to make?

2. My diagonal fov is 190 degrees. As mentioned in Car Surround view with fisheye cameras #793, this gives an error. Any way to fix this? Can the code be changed in any way for fov >= 180 degrees to work?(Any other alternative to using cameras with a lower fov?)

3. Any idea how to set soft_fm_config per my camera HW?

libXCam use OpenCV optical flow algorithm to do feature match between two stitching images. you can check the code by yourself for details.

This is not very helpful.

4. My outputs are horribly skewed even with the original viewpoints_range (64, 160), so I'm starting to think that my calibration files' format and parameters are wrong. Here is the command I run:
   ./test-surround-view --module soft --input ../front.yuv --input ../right.yuv --input ../back.yuv --input ../left.yuv --output output.yuv --in-w 640 --in-h 360 --out-w 1920 --out-h 640 --in-format yuv --fisheye-num 4 --cam-model camb4c1080p --blend-pyr-levels 1 --dewarp-mode bowl --scopic-mode mono --scale-mode dualcurve --frame-mode single --fm-mode none --fm-frames 1 --fm-status fmfirst --save true --loop 1

Following is a .zip file consisting of sample intrinsics and extrinsics files, the 4 input images and the output.
In the extrinsics file, I have skipped incorporating the extrinsics given by OCamCalib and have only put in the location of the cameras wrt the car dimensions for now.(Will add them in later.) Can you have a look?
libxcamdoubts.zip

[sandeshchand87]

@shreshtashetty Did you get the perfect output?