how to extract or parse the raw data? about gvins_recorder HOT 4 OPEN

You’re too kind,the app still works well.
I see it would takes a lot of time to improve the code(hope i didn't disrupt your schedule :) ),so please work on it according to your wish.

All the best with you!
Horry

from gvins_recorder.

Hi @belkahorry,

Thank you for trying out the sensor recorder.
Oh, I really did not provide any example of how to use the recorded data - that's embarrasing, lol. Sorry about that.

I have now written a tiny python script which loads up the recorded .json file, parses the data and visualizes it. It took me quite some time to realize that the base64 encoded images contain an additional header "data:image/png;base64," which needs to be stripped off when saving the images to disk.

I have updated the README with basic instructions and acknowledged your question here:
https://github.com/GSORF/GVINS_Recorder#example-of-recorded-sensor-data

I also wanted to upload my own example dataset used for plotting, however, I just noticed that I have recorded my face and therefore I better do some proper recording before uploading the data publicly... ;) I don't know when I will be able to, currently it is snowing outside, but once I get to it, I will upload a sample file as well.

Hope this is helpful to you, if you need anything else, feel free to ask.

Kind regards,
Adam

from gvins_recorder.

Hi Adam,
Much appreciate for your prompt reply and detailed explanation!
The script helps a lot and i‘ve got several clear results conveniently.

(I replaced the str 'video' with 'camera' in line 21 and 69,since my json file has no key ‘video’ )

There is one more question:
The extracted image's resolution is 300(W)x150(H) and is cropped 1/3 width on both side which are two transparent area now,only centre part remains(100(W)x150(H)) .

Is this default setting? or are there some compatibility problems with my device?
potrait:landscape:
screenshot:

Once again,thanks for providing the parsing code which really helped me out.

from gvins_recorder.

Hi @belkahorry,

sorry for my late answer.

Is this default setting? or are there some compatibility problems with my device?

Wow, you are right, thats a huge bug you have found here. It is neither a setting, nor a compatibility issue. It is just a bug, haha. The reason for this is, that I am using a HTML5 canvas to extract the image pixels from. This canvas obviously does (1) not scale with the phone orientation correctly, (2) is therefore wasting precious pixel space with whitespace and (3) it is waaaaay too small to actually do useful image processing on it (unless you have an awesome upsampling algorithm / AI). Thanks for pointing this out.

(I replaced the str 'video' with 'camera' in line 21 and 69,since my json file has no key ‘video’ )

Plus, you are also right about the parsing script. I don't know why my dataset had the video key, I think I had used an older dataset (actually, I had to reinstall the app on my phone, because I had a newer half-working-and-in-development build). I have updated the python script accordingly and also uploaded my own dataset which I have created today (see visualization below). Thanks for your kind help here once again.

Actually, as you have figured, I am very dissatisfied with the way how the camera and filesystem are handled / implemented at the moment. This is actually the main reason why there wasn't any update to my repo since 2021. I am using Cordova and HTML5 (which I love) for the implementation, but the performance is just not good enough for such processing-heavy tasks like recording sensor data in real-time, I believe. I have looked into alternative mobile development frameworks in the meantime and want to do a rewrite in one of those more modern ones from Facebook or Google. However, I just didn't have the time yet to improve the code.

Even though your issue with parsing seems to be fixed, let us leave this issue open for now. I think it is important for other users to see the limitations of the tool. Currently it is nothing more than a proof of concept. I am very sad, that the tool does not meet my own expectations for now and I will try to work on it in the future. Things are just a little stressful at the moment, so I cannot promise any date at this point. But I would like to sincerely thank you for investing your time to try out - and more importantly - give me constructive feedback on my tool.

Thank you and kind regards,
Adam

from gvins_recorder.