[le.li.zju]

Hi there,

I was trying to implement a camera live stream viewer, which reads the image with 'Cam2D_Snapshot' function and show it with 'cv::imshow'. My code is like

Code:

cam_item = RDK.Item(CAM_NAME)
bytes_img = RDK.Cam2D_Snapshot("", cam_item, "")
nparr = np.frombuffer(bytes_img, np.uint8)
img_socket = cv.imdecode(nparr, cv.IMREAD_COLOR)

But it turned out that 'Cam2D_Snapshot' function took around 200~400 ms, is there anything I can do to speed up the snapshot in order to increase the FPS?

[Sam]

The speed highly depends on the size of the image you are retrieving.
What is the resolution of the images?

[le.li.zju]

Forgot to mentioned the size of the image is 640*480

[Sam]

You should be able to have a 60 FPS feed with this resolution. Are you doing any other processing to the image in your Python code?

[le.li.zju]

Actually, it is the function "Cam2D_Snapshot" which is a bit slow, the processing of the images is not relevant to this issue.

I realized that having the preview window opened will help to reduce the execution time of "Cam2D_Snapshot" to 70~120 ms. But can it be even faster?

I have attached the .rdk file and the python script

  scene_with_camera.rdk (Size: 278.25 KB / Downloads: 267)

  server.py (Size: 1.47 KB / Downloads: 247)

You could run following commands to reproduce the results

> RoboDK scene_with_camera.rdk
> python3 server.py

[Sam]

For optimal results, undock the window, used fixed window size, and minimize the preview window.
You can do so in the API using:

Code:

    # Connect to the running instance of RoboDK (make sure it's already running)
    logging.info("Connecting...")
    RDK = Robolink()
    RDK.Connect()
    logging.info("Connected...")

    CAM_NAME = 'StaticCamera'
    static_camera = RDK.Item(CAM_NAME)

    # Optimal settings: undocked, minimized, fixed sensor size
    RDK.Cam2D_SetParams("SIZE=640x480 WINDOWFIXED MINIMIZED", static_camera)
    static_camera.setParam('Open', 1)

    for i in range(1,20):
        start_time = time.time()
        logging.info("Taking snapshot...")
        bytes_img = RDK.Cam2D_Snapshot("", static_camera)
        end_time = time.time()
        logging.info("Snapshot took %.6f seconds", (end_time - start_time))

I have around 12 to 25 ms.

More examples in the docs:
https://robodk.com/doc/en/PythonAPI/exam...#camera-2d

[le.li.zju]

For optimal results, undock the window, used fixed window size, and minimize the preview window.
You can do so in the API using:

Code:

    # Connect to the running instance of RoboDK (make sure it's already running)
    logging.info("Connecting...")
    RDK = Robolink()
    RDK.Connect()
    logging.info("Connected...")

    CAM_NAME = 'StaticCamera'
    static_camera = RDK.Item(CAM_NAME)

    # Optimal settings: undocked, minimized, fixed sensor size
    RDK.Cam2D_SetParams("SIZE=640x480 WINDOWFIXED MINIMIZED", static_camera)
    static_camera.setParam('Open', 1)

    for i in range(1,20):
        start_time = time.time()
        logging.info("Taking snapshot...")
        bytes_img = RDK.Cam2D_Snapshot("", static_camera)
        end_time = time.time()
        logging.info("Snapshot took %.6f seconds", (end_time - start_time))

I have around 12 to 25 ms.

More examples in the docs:
https://robodk.com/doc/en/PythonAPI/exam...#camera-2d

I have tried your approach and had FPS around 30Hz.

But I have got 2 questions still:

- By undocking the window, we would probably need to manually open the preview window, right? Is there a method in the API or parameters which we can set, to open the preview window automatically?

- And what exactly is "static_camera.setParam('Open', 1)" doing? Is it really necessary?

[Sam]

It seems you have answered your own question.

"static_camera.setParam('Open', 1)" opens the preview window.

[le.li.zju]

It seems you have answered your own question.

"static_camera.setParam('Open', 1)" opens the preview window.

Okkk, now it makes much more sense to me. So actually I should use "SetParam('Open', 1)" to open the preview window first and then call "RDK.Cam2D_Snapshot()" to set the parameters like image resolution, right?

I was following the python code snippet in the previous message (which called Cam2D_Snapshot first and then SetParam('Open', 1) ), and ended up only getting small image (100*75).