Tips and Trick: How I Vet 100 observations in less than 5 minutes

Quick screenshot during model training; currently sitting at 88% accuracy.

1 Like

My station (#4950) is mostly drowning in a sea of yellow, so this is extremely interesting to me!

This is running, it will vet and submit those where there is a high probability of a signal, or there is no signal, leaving you to vet and submit the observations where the model is uncertain.

kind regards,

PJM #91

1 Like

after 1 hour, im comparing , seen nothing change. is it still running?

currently at 168 of the backlog processed, with 67 confident (65 signal, 2 no_signal) and the remaining ~101 falling into “uncertain”. You happy for me to press the go button and submit these vettings ?

KR

PJM.

1 Like

In fact may be worth joining the SatNOGS IRC channel on libera.chat.

PJM

1 Like

May I suggest to use the Matrix option https://app.element.io/#/room/#satnogs:matrix.org as an old IRC user not my favorite but it is more stable then the irc ↔ matrix bridge.

Jan | PE0SAT

2 Likes

Please go ahead and submit them! Let’s see how the model works.

My vetting CV system is running against your obs; 653 real observations submitted successfully so far for station 4451 — 613 signal, 40 no_signal, 0 failed. Do you want me to process your obs every day ? So far my model has trained against 16280obs (10034 with signal / 6246 without signal ) 89% accuracy, pretty happy with that, at least as good as an average human vetter!

KR, Paul.

same with matrix?

yes. thank you!

1 Like

cool

for strong signal from terestrial like this should be red

I’ve scheduled your auto-vetting to run once per day, please keep an eye on it and let me know if something doesn’t look right. It will classify ‘with signal’ and ‘without signal’ and anything its unsure of will be left for you to vet, it will then learn from your human vetting to improve the model.

kind regards, Paul.

1 Like

ah okay. so your LLM will automatically learn from anything that unsure that left to me vetting. right? awesome

1 Like

Its more CV than LLM but yes it trains from human vetted obs, I’ve also added the other station 4950 so it will process mine #91 yours and his!

KR Paul.

2 Likes

Hi Paul,

Thank you for including my station (4950) in your automated vetting process! It’s an exciting approach, and I’m very glad to be a part of it. I’ll keep an eye on how it goes.

Best regards, Yukio (jp7dvx)

1 Like

Hi Paul,

Thanks again for including my station! While the yellow labels are still piling up, I caught this interesting one today.

Cheers, Yukio (jp7dvx)

this not signal from ridusat

1 Like