Capture csi camera with python

ps_nithin · September 5, 2025, 6:15am

Hi,

What is the recommended way to capture csi camera frames on jetson orin nano / jetson devices using python.

Thanks,

carolyuu · September 5, 2025, 6:33am

*** Please note that this reply is generated by LLM automatically ***

Capturing CSI camera frames on Jetson Orin Nano using Python involves several steps. Here’s a comprehensive guide to help you get started:

Prerequisites

Ensure you have a Jetson Orin Nano device with a CSI camera module connected.
Install the JetPack SDK on your Jetson Orin Nano device. This includes the NVIDIA Linux Driver Package, CUDA Toolkit, and other dependencies.
Install Python 3.x on your Jetson Orin Nano device.

Recommended Approach

Use the nvgstcapture application: This is a command-line tool provided by NVIDIA to capture CSI camera frames. You can use it to capture frames and save them to a file.

nvgstcapture-1.0 --help

This will display the available options for capturing frames.

Use the gst-python library: This is a Python wrapper for the GStreamer library, which is used by nvgstcapture. You can use this library to capture CSI camera frames in Python.

import gi
gi.require_version('Gst', '1.0')
from gi.repository import Gst

# Initialize GStreamer
Gst.init(None)

# Create a pipeline to capture CSI camera frames
pipeline = Gst.Pipeline()

# Add a CSI camera source to the pipeline
src = Gst.ElementFactory.make('nvcamerasrc', 'src')
pipeline.add(src)

# Add a video sink to the pipeline to display the frames
sink = Gst.ElementFactory.make('nvvidconv', 'sink')
pipeline.add(sink)

# Link the elements in the pipeline
src.link(sink)

# Set the pipeline to playing state
pipeline.set_state(Gst.State.PLAYING)

# Capture frames and save them to a file
while True:
    # Get the next frame from the pipeline
    buffer = pipeline.get_buffer()

    # Save the frame to a file
    with open('frame.jpg', 'wb') as f:
        f.write(buffer.data)

    # Display the frame using OpenCV (optional)
    import cv2
    frame = cv2.imdecode(buffer.data, cv2.IMREAD_COLOR)
    cv2.imshow('Frame', frame)
    cv2.waitKey(1)

This code captures CSI camera frames and saves them to a file named frame.jpg. You can modify the code to display the frames using OpenCV or perform other processing tasks.

Comparison with Other Approaches

Other approaches to capturing CSI camera frames on Jetson Orin Nano using Python include:

Using the pygst library: This is another Python wrapper for the GStreamer library. However, it is not as widely used as gst-python and may not be as well-maintained.
Using the opencv-python library: This library provides a Python interface to OpenCV, which can be used to capture CSI camera frames. However, it may not provide the same level of control as gst-python or nvgstcapture.

In summary, the recommended approach to capturing CSI camera frames on Jetson Orin Nano using Python is to use the gst-python library or the nvgstcapture application. These provide a robust and efficient way to capture frames and perform processing tasks.

*** Please give the thumbs up if you get answers you like. Or provide feedback to help us improve the answer. ***

ShaneCCC · September 5, 2025, 6:34am

Use the gst-python library: This is a Python wrapper for the GStreamer library, which is used by nvgstcapture. You can use this library to capture CSI camera frames in Python.

import gi
gi.require_version('Gst', '1.0')
from gi.repository import Gst

# Initialize GStreamer
Gst.init(None)

# Create a pipeline to capture CSI camera frames
pipeline = Gst.Pipeline()

# Add a CSI camera source to the pipeline
src = Gst.ElementFactory.make('nvcamerasrc', 'src')
pipeline.add(src)

# Add a video sink to the pipeline to display the frames
sink = Gst.ElementFactory.make('nvvidconv', 'sink')
pipeline.add(sink)

# Link the elements in the pipeline
src.link(sink)

# Set the pipeline to playing state
pipeline.set_state(Gst.State.PLAYING)

# Capture frames and save them to a file
while True:
    # Get the next frame from the pipeline
    buffer = pipeline.get_buffer()

    # Save the frame to a file
    with open('frame.jpg', 'wb') as f:
        f.write(buffer.data)

    # Display the frame using OpenCV (optional)
    import cv2
    frame = cv2.imdecode(buffer.data, cv2.IMREAD_COLOR)
    cv2.imshow('Frame', frame)
    cv2.waitKey(1)

This code captures CSI camera frames and saves them to a file named frame.jpg. You can modify the code to display the frames using OpenCV or perform other processing tasks.

Comparison with Other Approaches

Other approaches to capturing CSI camera frames on Jetson Orin Nano using Python include:

Using the pygst library: This is another Python wrapper for the GStreamer library. However, it is not as widely used as gst-python and may not be as well-maintained.
Using the opencv-python library: This library provides a Python interface to OpenCV, which can be used to capture CSI camera frames. However, it may not provide the same level of control as gst-python or nvgstcapture.

In summary, the recommended approach to capturing CSI camera frames on Jetson Orin Nano using Python is to use the gst-python library or the nvgstcapture application. These provide a robust and efficient way to capture frames and perform processing tasks.

ps_nithin · September 5, 2025, 11:44am

Hi,

I get the following error when i run the above code.

Traceback (most recent call last):
  File "/home/ps_nithin/cam.py", line 13, in <module>
    pipeline.add(src)
TypeError: Argument 1 does not allow None as a value

ps_nithin · September 6, 2025, 9:05am

Hi,

import gi
import numpy as np
import time

gi.require_version('Gst', '1.0')
from gi.repository import Gst, GObject, GLib
from PIL import Image

def on_new_sample(sink, data):
    sample = sink.emit("pull-sample")
    print("here")
    if sample:
        buffer = sample.get_buffer()
        caps = sample.get_caps()
        
        # Extract frame data and convert to OpenCV format
        # This part depends on the format in your pipeline (e.g., RGB)
        # For BGR, you can directly create a NumPy array
        
        # Example for RGB format:
        width = caps.get_structure(0).get_value("width")
        height = caps.get_structure(0).get_value("height")
        
        # Get raw buffer data
        success, mapinfo = buffer.map(Gst.MapFlags.READ)
        if success:
            frame_data = np.ndarray(
                (height, width, 3),  # Assuming RGB format
                buffer=mapinfo.data,
                dtype=np.uint8
            )
            buffer.unmap(mapinfo)
            
            # Process the frame
            Image.fromarray(frame_data).convert('RGB').save("gst.png")
        else:
            print("Failed to map buffer")
    return Gst.FlowReturn.OK

Gst.init(None)

pipeline_string = (
    "nvarguscamerasrc sensor-id=0 ! "  # Adjust sensor-id if multiple cameras
    "video/x-raw(memory:NVMM), width=(int)640, height=(int)480, format=(string)NV12, framerate=(fraction)30/1 ! "
    "nvvidconv ! "
    "video/x-raw, format=(string)BGRx ! "
    "videoconvert ! "
    "video/x-raw, format=(string)RGB ! "
    "appsink name=mysink"
)
pipeline = Gst.parse_launch(pipeline_string)

appsink = pipeline.get_by_name("mysink")
appsink.set_property("emit-signals", True)
appsink.connect("new-sample", on_new_sample, None) # Define on_new_sample function to process frames

pipeline.set_state(Gst.State.PLAYING)

loop = GLib.MainLoop()
try:
    loop.run()
except KeyboardInterrupt:
    pass
finally:
    pipeline.set_state(Gst.State.NULL)
    loop.quit()

The above code works but the output is not realtime. It is not giving the current frame. What could be the reason?

Thanks,

ps_nithin · September 6, 2025, 10:40am

The code works ok when frame processing is simple. But if i add time.sleep(3) after Image.fromarray(frame_data).convert('RGB').save("gst.png") the lag happens. How can this be fixed.

Thanks,

ps_nithin · September 6, 2025, 11:27am

Using

sample = appsink.emit(“try-pull-sample”, 100*Gst.MSECOND) helped.

nico.morua · September 17, 2025, 6:46pm

Hi @ps_nithin !

Hope you are doing well!

Recently I developed a Python application for a media server using gst-python and I documented it on this wiki:
Media Server with GStreamer: Python-based Video Streaming, Recording, and Snapshot Solution.

It includes an example with live streaming, snapshots, and recording — all in Python.
You can adapt it to your use case by modifying the main_pipeline_description to use your camera as the input.

I thought it might be helpful as you work on your own Python application.

Nico
Best regards
Embedded Software Engineer at ProventusNova

ps_nithin · September 18, 2025, 3:54am

Hi,

Thanks for sharing,

ps_nithin · September 18, 2025, 1:39pm

I worked out the code.

But some other things got broken. On jetson nano jp4.6.6 , gst-python runs without issues with python3.6. But i need python3.8 for numba. Numba runs in conda environment in which i am not able to install python3.6. I cant install numba outside conda with pip as it fails. Now i am left with not using gst-python.

Thanks,

system · October 8, 2025, 12:58am

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.

Topic		Replies	Views
Orin Nano IMX219 CSI Camera Python Jetson Orin Nano camera , python , jetson	4	52	November 16, 2025
Problem CSI camera Jetson Nano OpenCV Jetson Nano camera , opencv , gstreamer , python	3	5003	December 29, 2021
Jetson nano csi camera read opencv with python Jetson Nano camera , opencv	2	1677	July 23, 2022
CSI Camera not working Jetson Orin Nano Jetson Orin Nano camera	2	1750	November 24, 2023
How to capture an image of a CSI camera in python3? Jetson Nano camera	4	2017	February 13, 2024
Jetson Orinn Nano CSI camera unable to open camera Jetson Orin Nano camera	4	763	January 30, 2024
CSI Camera Capture Python3 - Jetson Orin Nano - select() timeout Jetson Orin Nano camera	4	357	February 10, 2025
Not able to get CSI camera feed into opencv-python Jetson Orin Nano camera	14	2255	August 1, 2023
Unable to run CSI camera on jetson nano suddenly Jetson Nano camera	2	664	March 21, 2023
Simple_Camera.py program not working on Orin Nano Jetson Orin Nano camera , jetson	4	96	August 27, 2025

Capture csi camera with python

Related topics