• DepthAI-v2Hardware

  • The workflow to undistort multiple color cameras on devices like OAK-FFC 3P/4P

I also tried to run the calibration for OAK-1-LITE-W with this json file, and it returned the same error expecting stereo_config.

  • erik replied to this.

    Hi lincolnxlw ,
    Could you create an issue on depthai github repo, so you get notified when this is fixed?

        5 days later

        Hi erik

        Thanks to your suggestion, the folks from the repo suggested a temporary fix. And I am able to do the calibration process. When I try to undistort the camera with Camera node according to the example from your PR, I am able to get a full undistorted image

        However, there are some features are not currently available in Camera node yet, so when I try to undistort with ColorCamera node following this example. I am only getting the "zoom in" version of the undistorted image.

        What I need to modify in order to get the full undistorted image using ColorCamera node?

        Thanks

        Lincoln

        • erik replied to this.

            Hi erik

            I will! I usually use depthai-library from docker for testing. But if I need to use depthai develop branch, does that mean I need to modify this docker file and build with that branch? Or I can find the tag in depthai-library that already built with the develop branch?

            • erik replied to this.

                Hi erik

                Thanks! How do I know which branch the tag is linked to? I was looking into this.

                    5 days later

                    Hi @erik,

                    Just want to give you an update with the "zoom-in" issue. Change to the latest develop branch didn't make a difference. But it can be fixed by adding a scaling factor to the camera intrinsics array while creating the maps for undistortion.

                    Here is the updated function getMesh function for this example

                    def getMesh(calibData: dai.CalibrationHandler, ispSize, focal_len_scaling_factor: float = 1.0):
    M1 = np.array(calibData.getCameraIntrinsics(camSocket, ispSize[0], ispSize[1]))
    d1 = np.array(calibData.getDistortionCoefficients(camSocket))
    R1 = np.identity(3)
    M1_new = M1.copy()
    M1_new[0][0] *= focal_len_scaling_factor
    M1_new[1][1] *= focal_len_scaling_factor
    mapX, mapY = cv2.initUndistortRectifyMap(M1, d1, R1, M1_new, ispSize, cv2.CV_32FC1)

    meshCellSize = 16
    mesh0 = []
    # Creates subsampled mesh which will be loaded on to device to undistort the image
    for y in range(mapX.shape[0] + 1): # iterating over height of the image
        if y % meshCellSize == 0:
            rowLeft = []
            for x in range(mapX.shape[1]): # iterating over width of the image
                if x % meshCellSize == 0:
                    if y == mapX.shape[0] and x == mapX.shape[1]:
                        rowLeft.append(mapX[y - 1, x - 1])
                        rowLeft.append(mapY[y - 1, x - 1])
                    elif y == mapX.shape[0]:
                        rowLeft.append(mapX[y - 1, x])
                        rowLeft.append(mapY[y - 1, x])
                    elif x == mapX.shape[1]:
                        rowLeft.append(mapX[y, x - 1])
                        rowLeft.append(mapY[y, x - 1])
                    else:
                        rowLeft.append(mapX[y, x])
                        rowLeft.append(mapY[y, x])
            if (mapX.shape[1] % meshCellSize) % 2 != 0:
                rowLeft.append(0)
                rowLeft.append(0)

            mesh0.append(rowLeft)

    mesh0 = np.array(mesh0)
    meshWidth = mesh0.shape[1] // 2
    meshHeight = mesh0.shape[0]
    mesh0.resize(meshWidth * meshHeight, 2)

    mesh = list(map(tuple, mesh0))

    return mesh, meshWidth, meshHeight

                    And here are the undistorted images with scaling factor=1.0, 0.8 and 0.6. Smaller the factor, the more "zoom out" the image is.

                    Thanks

                    Lincoln