RegulatedNoise - trade tool with zxctypo's best-in-class OCR (EliteOCRReader as-was)

One thing i would love to see is a rare trade run optimizer, ie look for rares and then tell you where you could sell them.
Also a button for current station would be nice, so i don't have to look it up in the list every time (and again, the list should be sorted by system first not the station name ;) , but that's just my opinion.)
Also say i want to gain rep in a certain station and maybe not be totally focused on the most profitable run at the moment, i would then want to trade as much as possible to help that system get the wares they need the most, could be a nice niche function of your tool.
 
the list should be sorted by system first not the station name

I had a look at this. Basically, changing *what you see*, namely "Station Name [System Name]" is going to be more work than it's worth at the moment. But changing the *order in which they appear* is simple enough that I could add it. So to be clear, you'd still see "Station Name [System Name]" in the various Station comboboxes, but the items in those comboboxes would be in System order.

That way at least stations in a system are grouped together, which might be close to what you want, even if it doesn't read as nicely as I'd like. Let me know what you think - obviously I'd hang this behaviour off a check box on the Settings tab I think we need at this point :)
 
Better idea. In the next version, the station combobox (and the to and from boxes in the station-to-station tab) will be ordered by distance from your current location. So your current system (or the "lightyears from" system) will always be at the top of the list, and the further you are from a system, the lower it will be in the list...
 
Hey, thanks for your hard work.

I am having a problem w/OCR accuracy.

I already had EliteOCRReader working flawlessly.

But when I use the OCR built into RegulatedNoise (on the exact same screenshots) I have large amounts of errors (Primarily with items that have a lot of N I or M next to each other).

Is this a calibration issue?
 
Last edited:
Hey, thanks for your hard work.

I am having a problem w/OCR accuracy.

I already had EliteOCRReader working flawlessly.

But when I use the OCR built into RegulatedNoise I have large amounts of errors (Primarily with items that have a lot of N I or M next to each other).

Is this a calibration issue?

Well, I guess so. If you can:
a) post an example Screenshot_????.bmp to imgur.com and link to it here, and
b) also attach Calibration.txt from your RN directory,
I'll take a look. Ideally, make sure that the Screenshot_????.bmp you choose, is one that gives you errors.
 
Hey, thanks for your hard work.

I am having a problem w/OCR accuracy.

I already had EliteOCRReader working flawlessly.

But when I use the OCR built into RegulatedNoise (on the exact same screenshots) I have large amounts of errors (Primarily with items that have a lot of N I or M next to each other).

Is this a calibration issue?

I also am getting lots of OCR errors. Increasing the brightness of the HUD in options seemed to help a bit but might be my imagination. It seems to also depend on what is going on in the background as the commodities screen is not completely opaque.
The most irritating thing about this is if it gets the station name wrong but nothing else (if it gets a commodity wrong it pauses) it can sometimes update the data before you have a chance to correct the field, or is there a setting so that you have to manually commit OCR readings?

Still loving the tool. :)
 
Well, I guess so. If you can:
a) post an example Screenshot_????.bmp to imgur.com and link to it here, and
b) also attach Calibration.txt from your RN directory,
I'll take a look. Ideally, make sure that the Screenshot_????.bmp you choose, is one that gives you errors.

Thanks... I'll get this when I am at home.

It won't be a problem, I get errors on all screenshots.
 
Increasing the brightness of the HUD in options seemed to help a bit but might be my imagination. It seems to also depend on what is going on in the background as the commodities screen is not completely opaque.

Yep, I'm sure both those things are true. If you wanted to go nuts, you could try modifying your install to use (say) bright yellow text, which would probably also increase the OCR hit rate.

Again though, post the screenshot and calibration.txt, and I can take a look and find a way forward :)

is there a setting so that you have to manually commit OCR readings?

Increasinly I think we need one. I'll make it the next feature if I can.

- - - - - Additional Content Posted / Auto Merge - - - - -

It won't be a problem, I get errors on all screenshots.

LOL! :)
 
I had a look at this. Basically, changing *what you see*, namely "Station Name [System Name]" is going to be more work than it's worth at the moment. But changing the *order in which they appear* is simple enough that I could add it. So to be clear, you'd still see "Station Name [System Name]" in the various Station comboboxes, but the items in those comboboxes would be in System order.

That way at least stations in a system are grouped together, which might be close to what you want, even if it doesn't read as nicely as I'd like. Let me know what you think - obviously I'd hang this behaviour off a check box on the Settings tab I think we need at this point :)

Better idea. In the next version, the station combobox (and the to and from boxes in the station-to-station tab) will be ordered by distance from your current location. So your current system (or the "lightyears from" system) will always be at the top of the list, and the further you are from a system, the lower it will be in the list...

I guess both are good way's to atleast fix it somewhat. Last night i was trying to find a route to trade for a certain stations.. but i gave up as i couldn't find the system :p
 
Any chance of a G19 LCD keyboard plugin small memory footprint , like to be able to see Rn in there Doubt the OCR would work, but the rest could
 
Compliments and many thanks for the tool and the work you put into it. :) So far things are working very well, I'm experiencing one recognition problem per three or four screenshots which is okay for me.
 
A couple of bugs ive experienced all on v1.41.

Bug1:
I was at 'sun takush' and grabbed the screens, traveled to 'LHS 3479' and tried to grab screenshots. Somehow it still thought i was at sun takush and updated sun takush prices. The current system from logfiles field was correctly updated. ive also tried to travel to other systems, but still just sun takush was updated.
I had to correct one OCR and i remember System Name in OCR Correction was set to sun takush, Ive tried to remove that but it didnt help.

Bug2:
A new clean install and now it works better, two systems was scanned without problem, then I came to Dalton Gateway. On first try the OCR Wrote D'alton and saved data on that. I tried again and then it got Oalton. I also had to do one correction, don't remember which, but with both wrong value and system name I tried to remove the correct value and push continue, as it says in the really useful tip. However the program crashes and if i open it, it just crash.
at line 806 in form1.cs the SystemLocations[] has count 0. exception is then ofc: The given key was not present in the dictionary.
I'm not familiar with your code but moving ImportSystemLocations(); above SetupGui(); in Public Form1() seem to solve that issue.
 
Why is it that EliteOCR can grab everything with 100% accuracy and without having to awkwardly calibrate it. As it stands now the EDDN is useless because of all the people feeding it useless info caught by RN.

I am not bashing the thing, btw. I still use it with CSVs grabbed with EliteOCR. If the screenshot thing worked as good as in that, I'd be singing praises to the heavens.
 
Why is it that EliteOCR can grab everything with 100% accuracy and without having to awkwardly calibrate it.

I did answer that question a few pages back... it's because it's a completely different program by a different person (and so on), and almost any similarity is coincidental. I'm not sure comparisons are very helpful here really.

As it stands now the EDDN is useless because of all the people feeding it useless info caught by RN.

Well, it should be entirely trivial for anyone maintaining an EDDN client to filter out such data. Maybe RN should be able to filter EDDN input too, even from itself (!). I note that *all* RN data is posted to the test schema, not the live schema, by default. But I also think that no matter how much work we put into filtering data going *into* EDDN, there will always be duff data, so it will always be the responsibility of the consumer application to clean that data to its liking.

Put another way, in my view at least, anyone looking for a "perfect" - or even near-perfect - dataset, shouldn't be using EDDN directly, but rather something based on and cleaned from EDDN. That's not to say we shouldn't upload the best data we can, but I don't personally think that's where the buck stops - and I say that as someone who maintains both an uploader and downloader. Just my view of it, others may disagree...?
 
As it stands , RN is working & well, the issues for many & myself is been able to correct any mistakes on the fly , I have mentioned before adding drop down box in the CORRECT VALUE , so everyone can have the correct commodity, spelling to choose from , the same with station names , system names. I noticing + at the end of names & ' in the middle of names
Idea to implement
Error checking on Prices
ie: IF price BUY/SELL is 12345 for commodity a & 123 for price SELL/BUY for same commodity a, then a flag is raised highlights the line for YOU to double check, those prices The screenshot used for the flag raised on the price discrepancy is saved separately , so it can be posted.

I did find last night, that RN had left off the last digit on only one commodity. It was obvious to see, as the price discrepancy between BUY SELL was huge , & no I don't have the screenshot , as I have delete checked.

The Delete screen shot option ,should first move/copy the screenshot to a temp folder first, to allow RN to do its error checking in the back ground while you fly to next station. When no discrepancy is found , screenshot is delete same for any discrepancy in names, call it the discrepancy temp folder.

What do you think ?
 
adding drop down box in the CORRECT VALUE

Wait - you mean in the "edit commodity" window, yes? in which case, i agree

the same with station names , system names

agree for station names - system names should be stable these days if coming from the logs (notwithstanding mrmp's post above)

I noticing + at the end of names & ' in the middle of names

agree, there's no filtering of this at present

IF price BUY/SELL is 12345 for commodity a & 123 for price SELL/BUY for same commodity

if the buy price is more than (say) 50% higher or lower than the sell price, do you mean? would make sense to me...

highlights the line for YOU to double check

in the Price Analysis tabs, so you can edit the data there, yes, and not before import?

It was obvious to see, as the price discrepancy between BUY SELL was huge , & no I don't have the screenshot , as I have delete checked.

We can delete after import as long as we help the user to spot these errors *before* import.

Also, you gave me what I think might be a really nice idea. On my system, at 2560x1440, the screenshots are about 10.5Mb each. On loading them into RN, pretty-much the first thing I do is pre-process them - render them black and white, tweak the contrast and brightness and so on. That produces a nice image for OCR'ing - it's the "trimmed" image in the bottom right. I just tried writing that out as a PNG, and it weighs in at 46Kb (!!). And it looks like this => http://imgur.com/UfnG3ew.png <= which I think you'll agree is pretty readable. If we're going to keep graphics related to previously-scanned images, it would be better (and easy) to save these and delete the original screenshots. (Obviously we need the station name too, that's easy enough too).

Comedy calculations: you can't really dock at more than one different station every five minutes, so if someone plays for eight hours a day, and takes five screenshots at each station, they will take 480 screenshots a day, or about 24Mb in black-and-white pre-processed PNG format - and that's got to be a big overestimate for virtually any real person. Makes me inclined to keep a configurable rolling number of days' worth of processed PNGs automatically so you can *always* go back and check old data, even if you and RN didn't spot it was dodgy at the time.

So many ideas, so little time! :(

- - - - - Additional Content Posted / Auto Merge - - - - -

moving ImportSystemLocations(); above SetupGui();

Absolutely correct, found that one myself last night. You can help me fix the rest of it too!! :D :p

I'm trying to find lists of station names to validate against, I think that's our best bet in that regard...
 
Last edited:
As it stands now the EDDN is useless because of all the people feeding it useless info caught by RN.

Yes it's an issues , would like the option that RN saves user imports screenshots & EDDN data collected into separate CSVs So while your using RN and running EDDN ,you can use the data or not, but if its all too much bad data, just reload YOUR save.

Given the Lightyears tab ,will only show the stations I'm around collecting data, away from others I can ignore EDDN as none of MY systems are been updated
BUT if your in a area full of people entering data & have EDDN enable then this option is pointless , hence the idea above.
 
Back
Top Bottom