Implementing a new type of camera in crappy

[Hyo-u]

Hi there,

Background:

Elbub supervisor here, first thing first, thank you for building Crappy, it clearly helps (at least our) small company that want to devellop their R&D.
The program built by Elbub on my specification works well (there is still sometime some error when reloading but nothing truly problematic).
But there are still some part to smooth up/implement/clean which will probably lead to another internship next year.
Disclamer : I don't usually code so take everything I say with a grain of salt (And even if I learned some python to follow Elbub internship, it is still very new to me).

What I want :

I have two cameras, I want them to take pictures (with a lower frequency than others inputs)
Store the time of when the picture is taken (to match with load and displacement) and their numbers in a file.

What has been done:

Tried : Take a picture with python (file : Test_camVimba)
Result : Work
Tried : use the camera with the open_cv module in crappy
Result : doesn't work
Reason possible : I think because open_cv doesn't work alone for accessing the camera
Tried : Create a custom camera class in crappy (file : Vimba.py/Camera_Vimba_0.1.py), add it to the crappy/camera/init.py and try to launch a camera block: Test_CamVimba_crappy.py)
Result : doesn't work
Reason possible : There should be some or a lot of mistakes in my code.

Question :

What I want to do doesn't seem that difficult on paper, but I can't wrap my head around how to add a camera to crappy.
Is there something obvious I doesn't see and have you any advice?

Info :

My camera : Alvium 1800 C-1620
Connector : USB3
OS : Windows 11
Python 3.10
Crappy 1.5.9
The Python librairy for my camera :
VimbaPython 1.2.1
Opencv 4.7.0.68
My files : Camera.zip

[WeisLeDocto]

Hello @Hyo-u,

Glad to hear that we can be of some help in your work !

For interfacing your camera with Crappy, I strongly suggest that you define a class in your test script (or in an other file imported in your test script), as shown here. In this case, there's no need to change anything in the source code of Crappy, including the __init__.py files. Hopefully the dedicated section in the documentation can be of some help to you.

I took a quick look at the vimba library that you want to use, it sems totally feasible to drive your camera from Crappy. If you're lost, I suggest you to modify your Test_camVimba.py script so that it doesn't include any context manager (i.e. with statement, hopefully Vimba allows it). Then you'll get a basic sequential script, and you'll be able to split the lines between the open, get_image and close methods of the Camera object. At first, avoid adding settings to your camera and keep your code as basic as possible.

Also, I'm not sure why you tried to use cv2.VideoCapture in the scripts that you sent, but if you can't read your camera with the built-in Camera_opencv of Crappy, then you'll have to rely on vimba only for grabbing the frame (like in your Test_camVimba.py example).

I don't have time now to dig in the vimba library, but I'll gladly help if you need further help with Crappy.
Weis

P.S. : Instead of just saying that your script doesn't work, please provide the Traceback (the console output) containing the error that you get. It will help understanding what's wrong with your code.

[Hyo-u]

Thank you for your detailed answer @WeisLeDocto ,

That is sort of what I have tried to do with my vimba.py. But I get error like this :

106, in __init__
    assert self.camera_name in camera_list or self.input_label,\
AssertionError: My_cam camera does not exist!

Ok good to know, modifying the __init__.py did cause some problem.

In the manual of Vimba Python, from what I understand it is specified to always use a with statement :

The Camera class implements a context manager. On entering the Camera's context, all camera features are detected and can be accessed only within the with statement (see pdf) .

from vimba import *
with Vimba.get_instance () as vimba:
cams = vimba.get_all_cameras ()

When I try to bypass it, I got this error :
RuntimeError: Called 'Vimba.__Impl.get_all_cameras()' outside of 'with' - statement scope.

From what I have read, Vimba use a GenICam transport layer (GenTL) to communicate, I will dig if I can find something useful without a with statement. If not I dunno what I can do.

PS : Yes it would have been more helpful, but I wasn't on the same PC as the one on which I tested my script. Will definitely do it when I come close to something.

Vimba_Python_Manual (1).pdf

[WeisLeDocto]

That is sort of what I have tried to do with my vimba.py. But I get error like this :

In the vimba.py example you define a class called 'Camera_Vimba', but when you then instantiate the Camera Block you pass the name 'My_cam' as an argument ! Hence the error that you get. You have to modify either of these names so that they match. Also, the first letter of the class must be capital and the others lowercase (will be changed in the next release).

Regarding the context manager issue, I get it that every call to Vimba's API has to be done in the context of a Vimba instance, and calls to Camera methods have to be done in the context of a Camera object. A Vimba instance is simply what is returned by Vimba.get_instance(), and is a regular Python object (see manual page 14). In a similar fashion, Vimba Camera objects are regular Python objects, so you can maybe try something like :

from vimba import *
vimba = Vimba.get_instance()
with vimba:
    cam = vimba.get_all_cameras()[0]
with vimba:
    with cam:
        frame = cam.get_frame()

Notice the two consecutive contexts of the same instance. If that succeeds, then you'll be able to split the calls between the several methods of a crappy.camera.Camera class, with each method its own context manager(s). If not, there are other possible workarounds but it might become less elegant.

Weis

[Hyo-u]

It is starting to work thanks to you (using your advices).
The camera is detected, Crappy launch now, tkinter open, and it loop the get_image function. I just see a black picture for now (I think, I still don't understand much but I'm getting there). I will probably be back with a lot more questions later, when I have more time to work on it.

I got an error that needed me to change a line in a crappy file : cameraConfig.py
from : fx = np.arange(0, out_size[1], out_size[1] / l0, dtype=np.float)
to : fx = np.arange(0, out_size[1], out_size[1] / l0, dtype=float)
With this error : AttributeError: module 'numpy' has no attribute 'float'.
np.float was a deprecated alias for the builtin float.

My code as of today :

import crappy
import numpy as np
from vimba import *
from time import time
import os

class Camera_vimba(crappy.camera.Camera):

    def __init__(self, resolution=(3040, 5312)):
        super().__init__()
        self.resolution = resolution
        self.frame = None
        self.name = "camera_vimba"
        self.add_setting('level', self._get_lvl, self._set_lvl, limits=(0, 255))

    def _get_lvl(self):
        return self.frame[0, 0]

    def _set_lvl(self, val):
        self.frame = np.ones(self.resolution, dtype=np.uint8) * val
        return val

    def open(self, **kwargs):
        global cam
        global vimba
        vimba = Vimba.get_instance()
        with vimba:
            cam = vimba.get_all_cameras()[0] 
            print(cam)
        self.frame = np.zeros(self.resolution, dtype=np.uint8)
        self.set_all(**kwargs)

    def get_image(self):
        with vimba:
            with cam:
                frame = cam.get_frame()
                print('got Frame')
        return time(), self.frame

    def close(self):
        print("Closing")
        del self.frame

if __name__ == "__main__":
    camera = crappy.blocks.Camera(camera='Camera_vimba', verbose=True, display_images=True, displayer_framerate=1)
    crappy.start()

[WeisLeDocto]

Good to hear you're making progress !

The deprecation error with numpy.float was normally fixed with commit 24715b2.

Took a quick look at your code : you could get rid of the resolution attribute and argument, as well as of the level setting and its getter and setter. The self.frame attribute is also not necessary. Also, instead of defining global variables, you should make them attributes of your class (using self.<name of attribute>).

The reason why you're getting a black image is because you acquire frame but return self.frame, these are not the same objects ! And one last thing, you'll probably have to use smth like np_frame = frame.as_numpy_ndarray() for converting the frame to the right format for Crappy to handle it.

Weis

[WeisLeDocto]

Closing discussion due to prolonged inactivity