Gen2 OCR Demo about depthai-experiments HOT 8 CLOSED

@dhruvsheth-ai seems like a good way to improve this.
I think that the root cause of flickering is that vstack appends whenever a new frame arrives as well as the text.
Maybe it be better to store all results first (of both transformed images and decoded text) and then display together once only.

If you'd like to give this a try in some form or another feel free to do so and open a PR. Will be glad to take a look:)

from depthai-experiments.

Sure! Will open a PR this week, with the mechanism you mentioned as well, will test and post the results

from depthai-experiments.

So, I tried using threading function to delay a specific part and not affect the rest, but that's not effective.
I'm trying to use another mechanism which you suggested

from depthai-experiments.

To open a PR later, with an effective threading solution

from depthai-experiments.

Is anyway this demo can be updated to run the inference for a provided image file?

Thanks

from depthai-experiments.

Is anyway this demo can be updated to run the inference for a provided image file?

Thanks

You can add arguments to the script to specify image input instead of Cam input. Check gen2-human-pose example to customize

from depthai-experiments.

@dhruvsheth-ai I am looking for an example where the pipeline input is an image instead of the camera. For ocr I would like to feed one image at a time. All the examples use camera stream as input to the pipeline and I would like to see some simple example where the input is just an image in the file system.

Thanks

from depthai-experiments.

I think we can add this as an alternate input.

@VanDavv - thoughts?

from depthai-experiments.