The 1MP, 2MP and 3MP direct the resolution of the camera on the device and really has nothing to do with the form's scanned resolution. This value directly impacts the performance of a DPX decode session. For example an employee badge or a form with a couple of barcodes would work great with a 1MP image, while a 3MP would take a much longer time as the fonts would be unnecessarily large. A driver's license for example is the same size as a badge but would need at least 2 or 3MP to cater to the busy-nature and font size in it. OMR fields would need a higher resolution in order to recognize them accurately.
In a previous version of the Template Builder, we did automatically use the uploaded image's resolution but soon realized that was debilitating as folks would load an incorrect [too low/high] resolution. Hence the move to have the user pick one of the 3 choices.
That said, I've noted down your confusion. Will try and alleviate this option setting either via some helper-pics or if possible try to automatically calculate it in the TB itself [with an option to change it manually].
For now, you could use the below as a helper :
1. 1 - 2 OCR/barcode fields : Ex: Employee badge, Barcode only forms with upto 3 barcodes : 1MP
2. OCR fields with mid sized fonts and/or upto 4 1D barcodes : 2 MP
3. Multiple OCR/OMR fields of varying sizes or really small fonts and/or combination of 1D and 2D barcodes. Ex: Most Drivers' License, Postal label in the showcase app , etc :3 MP
The device not recognizing the form is due to the fields marked as "Use field to identify the form". In a barcode only form, all the barcodes need to have this checkbox ticked, while in the mixed mode, only regions with an empty space around it [ex: logos, static text, etc] need this box ticked. it's recommended at least 3 fields on different sides of the forms be checked.
Also, the scanned image uploaded to the TB needs to be upright [not tilted or curved].
Please let me know if you have any further questions or still face issues. Share the sample pic and xml created so we can investigate.
Thanks for the explanation. I would suggest that the wording change as to ask those types of 'what are you really scanning' to help the software determine the best setting instead of how it is presented now.
Attached are the image (a real label scanned in using 300DPI on my scanner) and XML that I have not gotten to work correctly.
I am almost thinking that it may achieve better results if I do not use snapshot mode and pass in a picture taken with the camera. It seems like a lot of the times the preview snapshot mode is very clear to me, but does not allow the user to take the picture. Almost seems like auto focus is making the 'identification' process from not performing the best it can. Would it be quicker if the process was:
Take picture -> pass to engine -> see results
Viewfinder -> wait till engine decodes it> take picture -> see results
If the engine is sent a picture instead, does it behave differently (like try identifying the form by rotating on failure, etc)
DPX detects a large bounding box [ex, the dark border in your form], uses it as a reference frame to crop smaller regions of interest in it. The initial cropped image uploaded to the TB is presumed to be this initial reference frame. Alternatively, a barcode [a PDF417 for now] can be used as an anchor to get this larger reference frame as we know the barcode's size. This is achieved by the check box that says "Barcode's location is fixed".
Based on the image and template shared, there are a couple of different things that need to be rectified :
1. The template shows there's a barcode region but it's marked as "ANY Barcode". Given the form has two bar codes, this criteria is ambiguous and explains why the final output doesn't work.
Please specify the bar code symbology to discern between them if you wish to use it as an anchor. Alternatively, the dark border is sufficient to detect the border and the barcode as an anchor isn't really required. So you could just uncheck the "Barcode's location is fixed", in which case either of the bar codes decoded would be returned.
2. The template shows only this barcode to be output. There are 3 more fields [2 pic and 1 OCR]. These were all marked as "Use field to identify form" but the "Also read value from field" is not checked. Please check this if you want to output these values as well. Ignore this if you only meant to identify the barcode.
Also, please ensure the form is cropped to the proper dark borders when you upload it in the TB tool. As this maps to the "reference frame" on the device i.e. the software detects this boundary as it's starting point.
As for your concerns around preview and snapshot, we use the lower-resolution preview image for performance-sake. Barcodes and the form detection can be achieved using a low-res image while OCR/OMR etc need a sharper, high-res image achieved via a snapshot image.
Re: "Barcode's location is fixed"
I naturally answered yes to this question, because the barcode is always in the same location. But it sounds like the system interprets this option as 'use this barcode to identify the form'. Maybe this should reworded. Also, for the sake of the test I only 'care' about the barcode that I highlighted in TB. I expected the system to ignore anything that is not highlighted. I thought based on previous comments that I needed to tag 3 forms of identification.
Take into account that I have not received any training and if I did maybe I would not be asking these questions. But at the same time I think the system should be as super intuitive as possible. The proper setup of the form is critical in getting end user acceptance. Would be helpful if the TB template setup guided the user more:
1) User uploads image
2) TB system analyzes it from the start for 'acceptance' (clear image, identifiable areas, straightness, bounding box, etc whatever we look for) and either rejects it or gives it a rating 'green, red, yellow). if not a good candidate then gives suggestions for improving readability.
3) User is then guided via a wizard for template setup separating from identification from data collection
4) User is asked to select an area(s) of the form to help with identification and orientation (with suggested minimums and preferred). Like if I choose more 'use field to identify form' sections will it 'decode' faster or if I just pick a barcode for 'id' it will identify the form faster. This is not intuitive or clear at all.
5) User is then asked to select areas for data capture
6) User can block off areas to ignore (not sure if this helps with 'decode' time, but if blacking out sections of the image and mark as 'ignore' speeds things up then it would be good to have available.
Your inputs were duly noted. Will see what best we can do to improve this.
Worst case would a tutorial video / detailed help doc with images mitigate this issue ?