depthai-python: Problem running OAK-FFC-4P with Arducam 20MP IMX283 camera module
I am try to use OAK-FFC-4P with Arducam 20MP IMX283 Camera module. Running utilities/cam_test.py gives:
Enabled cameras:
rgb : color
left : mono
right : mono
camd : color
DepthAI version: 2.19.0.0
DepthAI path: /home/tem/.local/lib/python3.8/site-packages/depthai.cpython-38-x86_64-linux-gnu.so
[1844301021AE3D0E00] [3.3.3] [4.744] [ColorCamera(8)] [error] Camera not detected on socket: 3
[1844301021AE3D0E00] [3.3.3] [4.745] [MonoCamera(6)] [error] Camera not detected on socket: 2
[1844301021AE3D0E00] [3.3.3] [4.745] [ColorCamera(2)] [error] Camera not detected on socket: 0
[1844301021AE3D0E00] [3.3.3] [4.746] [MonoCamera(4)] [error] Camera not detected on socket: 1
Connected cameras:
USB speed: HIGH
IR drivers: []
Cam: rgb left right camd [host | capture timestamp]
FPS: 0.00| 0.00 0.00| 0.00 0.00| 0.00 0.00| 0.00
Running utilities/device_manager.py gives:
What is the problem here?
About this issue
- Original URL
- State: closed
- Created 2 years ago
- Comments: 27 (1 by maintainers)
Thanks @alex-luxonis Reinstall
depthai
solve the problem. It is working right now.@themarpe @cafemoloko Thanks for the update.
@tarekmuallim An initial version is available now, on the
imx283
branch, on top of the latest release. After checking out the branch, please runpython3 examples/install_requirements.py
Current status: streaming at 5312x3692 (maximum width supported now by the ISP processor is 5312), max FPS is 20. Exposure control is not working properly yet (applied exposure appears to be low at the default 20fps), and a lower FPS would apply a longer exposure. The ISO/gain setting works, however an observation with the Arducam module is that high gains create horizontal pattern artifacts. A workaround for now to get a higher exposure is to reduce the FPS, and/or increase the aperture of the lens, if available.
Needs a 4-lane camera socket on the FFC devices: RGB/CENTER/CAM_A, or CAM_D. 180-rotation doesn’t seem to be available with IMX283 (found only a vertical flip setting, and not horizontal mirror). To capture a normal upright image, the Arducam module has to be mounted with the FFC connector upwards.
Can use for a quick test:
python3 utilities/cam_test.py -cams rgb,c
and options like:-ds 2
to downscale 2 times (to 2656x1846),-fps 10.0
(can be lowered down to 0.074 – would apply very long exposures currently),-rs
to make the OpenCV window resizable (this option is not enabled by default as the OpenCV scaling may create artifacts), pressingC
to capture an image.By the end of the week we should get the exposure issue fixed, and will try to add more sensor resolution settings (cropping to 16:9 ratio approximately 5312x3000 , further cropping to 4K) - these should also work at a higher 30+ FPS.
Hi @themarpe We are working on developing an AI based barcode reader. For that we chose this sensor. We have already worked before with OAK-1 for another project “self checkout system” and the product is almost finished and we are happy with the results. For that we chose OAK-FFC-4P. We are using c++ “DepthAI Core” for development and python for some quick tests.
UVC mode is not enough as the deep learning model and the computation should be onboard.
We have contacted ArduCAM and they said that the IMX283 camera module is released only to be used with OAK! But as I understand from you it is not supported and looks like there is no plan to support it!
At the beginning I thought that this was some sort of a bug, for that I reported on github. Do I have to contact support@luxonis.com ?