frame.get_status() returns FrameStatus.Incomplete

[guanggevwu]

Hi,

I have some typical code:

def handler_one_by_one(self, cam: Camera, stream: Stream, frame: Frame):
    if frame.get_status() == FrameStatus.Complete:
        self.logger.info('Frame acquired: {}'.format(frame))
        frame_array = np.squeeze(frame.as_numpy_ndarray())
        self.imageq.put(frame_array)
    self.camera.queue_frame(frame)

if self._trigger_source in ['Software', 'External']:
    self.camera.start_streaming(self.handler_one_by_one)
    self.logger.info(
        f'Ready to receive {self._trigger_source} triggers')

I noticed sometimes the images were skipped and were not added to the queue (self.imageq). It turned out frame.get_status() returned FrameStatus.Incomplete sometimes. But I want to obtain all the images and save them. One reason is possibly that my bandwidth is not large enough since I have lots of other cameras in my network. I am willing to wait for a few seconds or even longer time to acquire the one image triggerred by software trigger or external trigger. What should I do?

[John-AlliedVision]

In the first instance you need to make sure you are getting good bandwidth in general. E.g. Items such as Jumbo packets are set.

I guess, secondly, you need to have a look at how close to (or over the edge) you are with respect to bandwidth at the moment. Vimba X Viewer, part of the SDK, allows a useful live view of the Stream Statistics so that you can see, for a particular camera, whether packets are being resent or frames dropped. If multiple cameras are seeing resent packets or dropped frames then you may need to reduce the DeviceLinkThroughputLimit or frame rates for some or all of the cameras. Resent packets will make any bandwidth limitation worse.

Note also, in the Alvium Features Reference, there are some features in the settings section (page 304 ff) that give some control over the way the SDK interacts with the camera to retrieve packets. For example, GVSPMaxRequests, GVSPTimeout, GVSPMissingSize et. al.. Each of these features has a description so that you can see what it is trying to do. They may be able to help if there are temporary periods of bandwidth overload but, if it tuns out that all the other cameras are consuming all the bandwidth, you may have to step back the throughput of some or all the other cameras.

Thank you and best wishes.

[guanggevwu]

I am using the default Jumbo packet size as 1500. I know a larger size may result a faster transfer.
I have two cases. (1) First is that I have many cameras (not only Allied Vision cameras) and once they receive triggers at the same time, my data acquisiton script will try to get the images. In this situation, the bandwidht is temperarily limited. (2) Second is if I have a camera running in live mode at a high fps all the time, all available bandwidth are almost always consumed. However, I guess it should still be doable to acquire one image from the Manta cameras, as long as I wish to wait for tens of second.
I will try these GVSP feature to see if I can get a good result.