[BUG] Cameras on different subnets cannot be launched

[jbarry-bdai]

Cameras on different subnets/LANs cannot be launched even when the IP is specified. I believe this is because on lines 199-202 of Camera.cpp, a call is made to getAllAvailableDevices and an error is thrown if that doesn't return devices before it checks whether there's a camera on the specified IP:

                std::vector<dai::DeviceInfo> availableDevices = dai::Device::getAllAvailableDevices();
                if(availableDevices.size() == 0) {
                    throw std::runtime_error("No devices detected!");
                }

According to the documentation, this is expected behavior when the camera is on a different LAN. Therefore, I think the code needs to check if the user-provided IP works even if devices are not returned from getAllAvailableDevices.

Minimal Reproducible Example

1. Connect to an OAK-D camera over Ethernet but set your netmask to 255.255.0.0 (instead of 255.255.255.0).
2. Confirm you can still ping the camera's IP address
3. Set camera/i_ip in the config file to the camera's IP address
4. Try to launch ros2 launch depthai_ros_driver camera.launch.py - it will not find the device even though it's pingable

However, if we use the depthai SDK directly, we can use the camera:

from depthai_sdk import OakCamera
# oak = OakCamera() <- This will fail because it's not on the same subnet
oak = OakCamera(device="<IP address>")  # This will succeed
color = oak.create_camera('color')
# Visualize the color camera to prove we're really connected
oak.visualize([color])
oak.start(blocking=True)

Expected behavior
When the camera IP is set, the launch file should be able to find the camera and launch successfully even when not on the same subnet.

[Serafadam]

Hi, thanks for the report, a bugfix will be available for testing this week.

[Serafadam]

Hi, sorry for the delay, could you test if this branch works for you?