Reopen camera after device.close() throws segmentation fault

Bbgro82 · Jul 15, 2023

My code is quite long, but essentially I am running device.close() on depthai.Device. After this I try to reopen the camera but hit a hard error when I call device.getConnectedCameras(). Not sure what if I am doing something wrong here or what the issue is, would appreciate any suggestions or advice. I am running on a raspberry pi zero armv6. Thank you!

Error details:

Stack trace (most recent call last):
#5    Object "/home/pi/.local/lib/python3.9/site-packages/depthai.cpython-39-arm-linux-gnueabihf.so", at 0xb569ddeb, in
#4    Object "/home/pi/.local/lib/python3.9/site-packages/depthai.cpython-39-arm-linux-gnueabihf.so", at 0xb56df557, in
#3    Object "/home/pi/.local/lib/python3.9/site-packages/depthai.cpython-39-arm-linux-gnueabihf.so", at 0xb5972f93, in dai:eviceBase::getConnectedCameras()
#2    Object "/home/pi/.local/lib/python3.9/site-packages/depthai.cpython-39-arm-linux-gnueabihf.so", at 0xb59985a7, in nanorpc::core::client<nanorpc::packer::nlohmann_msgpack>::result nanorpc::core::client<nanorpc::packer::nlohmann_msg
pack>::call<>(unsigned long long)
#1    Object "/lib/arm-linux-gnueabihf/libc.so.6", at 0xb6cff90f, in
#0    Object "/home/pi/.local/lib/python3.9/site-packages/depthai.cpython-39-arm-linux-gnueabihf.so", at 0xb5a434b3, in backward::SignalHandling::sig_handler(int, siginfo_t*, void*)
Segmentation fault (Address not mapped to object [0x8])
Segmentation fault

jakaskerl · Jul 16, 2023

Hi bgro82
You will have to wait some time before connecting to your camera again. Also make sure you are running the latest depthai version.

Thanks,
Jaka

Bbgro82 · Jul 22, 2023

Thanks, I am running depthai 2.22.0.0 and am still hitting the same error even with waiting up to 4 minutes between closing the camera and attempting to reopen. Is there a set amount of time I should be trying to wait? Is there anything else I can/should do before I try to reopen the camera?

jakaskerl · Jul 22, 2023

Hi bgro82
I was thinking more 10 seconds max, so the camera not being able to connect after 4 minute is a different problem then. Can you shave down the code as much as possible for the error to still occur and paste the code below? Keep as little functionality as you can.

Thanks,
Jaka

Bbgro82 · Jul 24, 2023

sure thing, just culled things down and verified this still throws the same error. It wouldn't let me upload the .py file here but let me know if easier to send to you somehow.

import time
import depthai as dai
import blobconverter

def create_pipeline(depth):
syncNN = True

# Start defining a pipeline
pipeline = dai.Pipeline()
# pipeline.setOpenVINOVersion(version=dai.OpenVINO.Version.VERSION_2021_4)
# Define a source - color camera
colorCam = pipeline.create(dai.node.ColorCamera)

# Define source and output for system info (temps/cpu)
sysLog = pipeline.create(dai.node.SystemLogger)
linkOut = pipeline.create(dai.node.XLinkOut)

linkOut.setStreamName("sysinfo")

# set system info pipeline to 1Hz sample rate
sysLog.setRate(1)

# Link
sysLog.out.link(linkOut.input)

if depth:
mobilenet = pipeline.create(dai.node.MobileNetSpatialDetectionNetwork)
monoLeft = pipeline.create(dai.node.MonoCamera)
monoRight = pipeline.create(dai.node.MonoCamera)
stereo = pipeline.create(dai.node.StereoDepth)
else:
mobilenet = pipeline.create(dai.node.MobileNetDetectionNetwork)

colorCam.setPreviewSize(512, 512)
colorCam.setResolution(dai.ColorCameraProperties.SensorResolution.THE_1080_P)
colorCam.setInterleaved(False)
colorCam.setColorOrder(dai.ColorCameraProperties.ColorOrder.BGR)

mobilenet.setBlobPath(blobconverter.from_zoo("person-vehicle-bike-detection-crossroad-1016", shaves=6, version="2022.1"))

mobilenet.setConfidenceThreshold(0.5)
mobilenet.input.setBlocking(False)

if depth:
monoLeft.setResolution(dai.MonoCameraProperties.SensorResolution.THE_400_P)
monoLeft.setBoardSocket(dai.CameraBoardSocket.CAM_B)
monoRight.setResolution(dai.MonoCameraProperties.SensorResolution.THE_400_P)
monoRight.setBoardSocket(dai.CameraBoardSocket.CAM_C)

# Setting node configs
stereo.initialConfig.setConfidenceThreshold(255)
stereo.depth.link(mobilenet.inputDepth)
stereo.setDepthAlign(dai.CameraBoardSocket.CAM_A)

mobilenet.setBoundingBoxScaleFactor(0.5)
mobilenet.setDepthLowerThreshold(100)
mobilenet.setDepthUpperThreshold(5000)

monoLeft.out.link(stereo.left)
monoRight.out.link(stereo.right)

xoutRgb = pipeline.create(dai.node.XLinkOut)
xoutRgb.setStreamName("rgb")
colorCam.preview.link(mobilenet.input)

if syncNN:
mobilenet.passthrough.link(xoutRgb.input)
else:
colorCam.preview.link(xoutRgb.input)

xoutNN = pipeline.create(dai.node.XLinkOut)
xoutNN.setStreamName("detections")
mobilenet.out.link(xoutNN.input)

return pipeline

with dai.Device() as camera_device:

while True:
print("Initiating camera")
# initiate camera and pipeline
cams = camera_device.getConnectedCameras()
depth_enabled = dai.CameraBoardSocket.CAM_B in cams and dai.CameraBoardSocket.CAM_C in cams

# Start pipeline
camera_device.startPipeline(create_pipeline(depth_enabled))

# Output queues will be used to get the rgb frames and nn data from the outputs defined above
previewQueue = camera_device.getOutputQueue(name="rgb", maxSize=1, blocking=False)
detectionNNQueue = camera_device.getOutputQueue(name="detections", maxSize=1, blocking=False)
sysinfo_queue = camera_device.getOutputQueue(name="sysinfo", maxSize=1, blocking=False)

frame = None
detections = []
frame_counter = 0
color = (255, 255, 255)

# camera loop
print("Starting camera loop for 200 iterations")
while frame_counter < 200:
inPreview = previewQueue.get()
frame_stereo = inPreview.getCvFrame()

inNN = detectionNNQueue.get()
detections = inNN.detections

frame_counter += 1

print("Closing camera")
camera_device.close()

time.sleep(120)

jakaskerl · Jul 25, 2023

Hi bgro82
What I changed: you had camera_device outside your loop. When ending first iteration, this would close the device, making it unavailable in the next iteration.

Updated script:

import time
import depthai as dai
import blobconverter

def create_pipeline(depth):
    syncNN = True

    # Start defining a pipeline
    pipeline = dai.Pipeline()
    # pipeline.setOpenVINOVersion(version=dai.OpenVINO.Version.VERSION_2021_4)
    # Define a source - color camera
    colorCam = pipeline.create(dai.node.ColorCamera)

    # Define source and output for system info (temps/cpu)
    sysLog = pipeline.create(dai.node.SystemLogger)
    linkOut = pipeline.create(dai.node.XLinkOut)

    linkOut.setStreamName("sysinfo")

    # set system info pipeline to 1Hz sample rate
    sysLog.setRate(1)

    # Link
    sysLog.out.link(linkOut.input)

    if depth:
        mobilenet = pipeline.create(dai.node.MobileNetSpatialDetectionNetwork)
        monoLeft = pipeline.create(dai.node.MonoCamera)
        monoRight = pipeline.create(dai.node.MonoCamera)
        stereo = pipeline.create(dai.node.StereoDepth)
    else:
        mobilenet = pipeline.create(dai.node.MobileNetDetectionNetwork)

    colorCam.setPreviewSize(512, 512)
    colorCam.setResolution(dai.ColorCameraProperties.SensorResolution.THE_1080_P)
    colorCam.setInterleaved(False)
    colorCam.setColorOrder(dai.ColorCameraProperties.ColorOrder.BGR)

    mobilenet.setBlobPath(blobconverter.from_zoo("person-vehicle-bike-detection-crossroad-1016", shaves=6, version="2022.1"))

    mobilenet.setConfidenceThreshold(0.5)
    mobilenet.input.setBlocking(False)

    if depth:
        monoLeft.setResolution(dai.MonoCameraProperties.SensorResolution.THE_400_P)
        monoLeft.setBoardSocket(dai.CameraBoardSocket.CAM_B)
        monoRight.setResolution(dai.MonoCameraProperties.SensorResolution.THE_400_P)
        monoRight.setBoardSocket(dai.CameraBoardSocket.CAM_C)

        # Setting node configs
        stereo.initialConfig.setConfidenceThreshold(255)
        stereo.depth.link(mobilenet.inputDepth)
        stereo.setDepthAlign(dai.CameraBoardSocket.CAM_A)

        mobilenet.setBoundingBoxScaleFactor(0.5)
        mobilenet.setDepthLowerThreshold(100)
        mobilenet.setDepthUpperThreshold(5000)

        monoLeft.out.link(stereo.left)
        monoRight.out.link(stereo.right)

    xoutRgb = pipeline.create(dai.node.XLinkOut)
    xoutRgb.setStreamName("rgb")
    colorCam.preview.link(mobilenet.input)

    if syncNN:
        mobilenet.passthrough.link(xoutRgb.input)
    else:
        colorCam.preview.link(xoutRgb.input)

    xoutNN = pipeline.create(dai.node.XLinkOut)
    xoutNN.setStreamName("detections")
    mobilenet.out.link(xoutNN.input)

    return pipeline

if __name__ == "__main__":

    while True:
        camera_device = dai.Device(create_pipeline(True))
        print("Initiating camera")
        # initiate camera and pipeline
        
        cams = camera_device.getConnectedCameras()
        
        depth_enabled = dai.CameraBoardSocket.CAM_B in cams and dai.CameraBoardSocket.CAM_C in cams
        # Start pipeline

        # Output queues will be used to get the rgb frames and nn data from the outputs defined above
        previewQueue = camera_device.getOutputQueue(name="rgb", maxSize=1, blocking=False)
        detectionNNQueue = camera_device.getOutputQueue(name="detections", maxSize=1, blocking=False)
        sysinfo_queue = camera_device.getOutputQueue(name="sysinfo", maxSize=1, blocking=False)

        frame = None
        detections = []
        frame_counter = 0
        color = (255, 255, 255)

        # camera loop
        print("Starting camera loop for 200 iterations")
        while frame_counter < 200:
            inPreview = previewQueue.get()
            frame_stereo = inPreview.getCvFrame()

            inNN = detectionNNQueue.get()
            detections = inNN.detections

            frame_counter += 1

        print("Closing camera")
        camera_device.close()

        time.sleep(10)

Thanks,
Jaka

Bbgro82 · Jul 25, 2023

oh man, so simple, I should have realized that. I appreciate you taking the time to help me solve that, thank you!

Bbgro82 · Jul 31, 2023

@jakaskerl this question is not directly related to my earlier one. But is it possible to run the same code in a "passthrough" type of state only where only RGB frames are pulled from the camera but no NN is running? Not looking to have you write my code for me but any suggestions would be appreciated. I've tried excluding everything with mobilenet and xoutNN but that still gives some errors: "RuntimeError: StereoDepth(5) - No output of StereoDepth is connected/used!" I'm not sure if what I'm trying to do just is not possible or if I'm going about it in the wrong way.

jakaskerl · Aug 1, 2023

Hi bgro82
Yes, it's possible, but you will have to add under IF block anything to do with NN's side of the pipeline. If you only want rgb, then you have to make sure you don't create a StereoDepth node --> this will result in the error you are experiencing.

Remove the red part, and directly link colorcamera to xlinkout. You should be able to achieve this with normal IF statements.

Thanks,
Jaka

Bbgro82 · Aug 1, 2023

Amazing, figured it out, thank you so much!