Connecting the Arducam Quad 12MP (IMX477) Color Camera
language
eng
date
Dec 29, 2025
slug
arducamQuadIMX477
author
status
Public
tags
Arducam
Embedded Vision
IMX477
Camera
summary
An overview of the Arducam Quad 12MP (IMX477) color camera and how to connect it.
type
Post
thumbnail
category
Embedded Vision
updatedAt
Jan 4, 2026 04:54 AM
Connection Guide
Arducam synchronizes sensor-board cameras using its own method (e.g., different clocking), so you cannot connect other embedded cameras to the Multi Adapter board. You must use the Arducam Camarray HAT + Arducam cameras together.
More details about the Arducam Quad 12MP (IMX477) Color camera:
The Camarray HAT setup differs depending on which embedded computer you use.
This test was done on a Raspberry Pi 4 with Bullseye OS.
- Use the built-in camera driver on Raspberry Pi 4 (Bullseye)
sudo nano /boot/config.txt #Find the line: camera_auto_detect=1, update it to: camera_auto_detect=0 dtoverlay=imx477 #Save and reboot.
- Control the Camarray HAT via I2C
(Select how many cameras to combine / how the images are stitched)
At first I couldn’t get I2C to show up no matter what I tried, and I wasted a lot of time—turns out the flat cable wasn’t fully seated. Make sure it’s inserted deeply and properly. Surprisingly, there isn’t much information about this online.
# Set to single channel 0 i2cset -y 10 0x24 0x24 0x02 # Set to single channel 1 i2cset -y 10 0x24 0x24 0x12 # Set to single channel 2 i2cset -y 10 0x24 0x24 0x22 # Set to single channel 3 i2cset -y 10 0x24 0x24 0x32 # Set to double channel (single channel 0 and single channel 1) i2cset -y 10 0x24 0x24 0x01 # Set to double channel (single channel 2 and single channel 3) i2cset -y 10 0x24 0x24 0x11 # Set to four in one mode (Default) i2cset -y 10 0x24 0x24 0x00
- Camera control - Libcamera
- Installation (Quick Start - Arducam Wiki )
- How to use built in app(libcamera and libcamera-apps - Arducam Wiki)
- Libcamera C++ example(GitHub - edward-ardu/libcamera-cpp-demo: libcamera c++ demo )
- Streaming using GStreamer(pi to PC)
Troubleshooting
V4L2 driver for arducam IMX477.
To confirm the driver is detected correctly. If you see something like this, the driver is loaded
jamesyeon@raspberrypi:~ $ dmesg | grep -E "imx477|imx219|arducam"~ [ 8.425914] imx477 10-001a: Device found is imx477 [ 8.426711] imx477 10-001a: Consider updating driver imx477 to match on endpoints
Cannot Allocate Memory
sudo nano /boot/cmdline.txt
Add cma=400M at the very end:
console=serial0,115200 console=tty1 root=PARTUUID=6ad9d548-02 rootfstype=ext4 fsck.repair=yes rootwait quiet splash plymouth.ignore-serial-consoles cma=400M
Test code
I extended it with features like zoom in/out and saving images. (Code is available in the referenced file in my repo/post.)
Observed limitations of Arducam + commercial embedded cameras
1) Camera cable length limitation
MIPI cable max length is commonly considered < 0.3 m. In practice, typical off-the-shelf setups usually need ≤ 0.2 m cables.
To fill the target FOV without severe barrel distortion, I needed a working distance (WD) of about 0.1 m. With this mechanical structure, it’s not easy to place four cameras so they can view one sample from four directions (the board tends to end up above or below the sample).
2) Need for an industrial embedded board
- Common boards like Jetson Nano or Raspberry Pi are basically evaluation kits: a CPU chip plus lots of I/O for development.
- For real applications, you usually need a custom SoM (System-on-Module) design using the target chip.
- (Reference: AMD-Xilinx materials)
3) Low customer accessibility to Linux development (OS/version fragmentation)
- Many embedded / machine-vision camera makers rely on kernel drivers provided by Linux for camera communication and image acquisition.
- Because drivers vary across Linux distributions and versions (Ubuntu vs Debian, etc.), you must confirm support and build your environment accordingly.
4) Limit on number of cameras per board
- With Arducam Camarray HAT, the practical maximum is 4 cameras per board.
- It’s difficult to increase this. Even with 4 cameras, the system typically acquires frames sequentially and the Camarray HAT stitches 12MP images in memory and outputs them as if they were from a single camera.
- To increase or decrease camera count, hardware modification is required.
5) Difficulty delivering result data externally
Likely customer needs fall into two categories:
- Original image O + defect detection data O
- Original image X + defect detection data O
There’s no ready-made platform/protocol to deliver both images and defect data to a customer PC. So you’d likely need to send packets via socket communication.
The challenge: like with GigE Vision cameras, you need a way to validate transmission integrity → a proper protocol is needed.
6) Hard to synchronize capture with external devices (lighting/motion)
Even if you use Raspberry Pi / Jetson Nano I/O, Arducam models that support external triggers are limited.
Streaming with GStreamer
On Raspberry Pi 4 (64-bit Bullseye), GStreamer 1.18.4-2.1 is installed by default. Verify with:
dpkg -l | grep gstream
This effectively means Debian 11 Bullseye is installed.
# install a missing dependency $ sudo apt-get install libx264-dev libjpeg-dev # install the remaining plugins $ sudo apt-get install libgstreamer1.0-dev \ libgstreamer-plugins-base1.0-dev \ libgstreamer-plugins-bad1.0-dev \ gstreamer1.0-plugins-ugly \ gstreamer1.0-tools \ gstreamer1.0-gl \ gstreamer1.0-gtk3 # if you have Qt5 install this plugin $ sudo apt-get install gstreamer1.0-qt5 # install if you want to work with audio $ sudo apt-get install gstreamer1.0-pulseaudio
gst-launch-1.0 libcamerasrc ! video/x-raw, width=640, height=480, framerate=30/1 ! videoconvert ! videoscale ! clockoverlay time-format="%D %H:%M:%S" ! autovideosink ! tcpserversink port=8888 host=localhost
gst-launch-1.0 -v azurekinectsrc ! videoconvert ! x264enc key-int-max=12 byte-stream=true ! mpegtsmux
UDP streaming example (image streaming)
gst-launch-1.0 -v libcamerasrc ! video/x-raw,width=640,height=480 ! videoconvert ! jpegenc ! rtpjpegpay ! udpsink host=172.18.1.0 port=5000
gst-launch-1.0 udpsrc port=5000 ! application/x-rtp,encoding-name=JPEG,payload=26 ! rtpjpegdepay ! jpegdec ! autovideosink
Cable extension attempt
[CSI-HDMI 케이블 확장 모듈](Arducam CSI-HDMI 케이블 확장 모듈 [B0091] / 디바이스마트 (devicemart.co.kr))
- Introduction page : MIPI CSI 2 to HDMI Converter Board & Extension
However, after researching and purchasing, I found out that CSI-to-HDMI extension does NOT support IMX477. The manufacturer said it was a labeling/specification mistake.
If you plan to buy this, confirm IMX477 support with the vendor first before purchasing.
Related posts
Axelera AI's large investment attraction: Reasons why European AI semiconductors are gaining attention again
Mar 9, 2026
Dutch AI semiconductor startup Axelera AI has secured an investment of $250 million. This news goes beyond simple funding; it reaffirms the presence of European AI semiconductor startups and the growth potential of the edge semiconductor market for AI inference. I summarized why Axelera AI is gaining attention and what this investment means for the industrial AI and machine vision sectors.
GenFeA: Lowering Barriers to Machine Vision Integration
Jan 11, 2026
The EMVA GenIcam group presented GenFeA, which makes camera integration easier in embedded and heterogeneous architectures. Let's take a look at the details and prospects.
Computer Vision in the World Cup: From Goal-Line Decisions to Semi-Automated Offside
Jul 6, 2026
World Cup officiating technology is moving beyond simple video review and turning stadiums into real-time data spaces. This post explains how goal-line technology, semi-automated offside, connected-ball data, and optical tracking systems help improve both fairness and the flow of the game.