Coming Soon
Scales
Notes:
ESP8266, ESP32, Raspberry Pi, chromebook?
HX711 for measuring load cells
74HC4051 multiplexer (used to take 1 signal and multiply (for hacking existing signals)
50kg load cells x4 / scale
variants
Fishing scale (tap into signal, push to esp8266. may need hx711 and multiplexer
Floor scales. same as above. Minor update the scales use bluetooth/wifi. see if I can pair a device to read the measurements instead of modifying the scale itself.
Homemade. 4x 50kg load cells hx711 esp8266, lcd1602 screens with drivers.
Receipt Scanning
So far after research Epson ES-50 is the recommended scanner. It is one of the lowest cost on the market, is available on all platforms (mac,pc,unix,pi), and has SANE support. Avoid Brother scanners, Near Receipts, and some epsons like the ds-620,ds-640, ds-30, ds-40 which are common [but ds-70 is ok]). Refer to this website if you want to do your own research. I recommend a scanner that is on the SANE list . Here
Notes.
A. OCR API somewhere - so far so good, this seems to be working
B. Use Pi with Tesseract, which is what most free online ones use. This is more favorable as the PI has hdmi , for other things like showing stats.
late night sleeping pill note: could use multi-passes to determine information and/or verify.. like use tesseract on pi, then a free one, then maybe google.
C. Google Cloud Document AI - in progress [3 processors in testing]***
D. forgot, chromebook??
current note to self, check amazon ocr options.
*** okay, the free credits they give, are trial only. I thought the credit you get was monthly, and anything over had a fee.. but it expires. just watch out, you may get billed for usage.
Random notes.
aspell
tesseract-ocr
Flow - draft.
Person comes in. items on scale, lock the weight. a button should execute a query to send the data to the central db.
receipt is put in a scanner to capture.
AI should do the following
get store name, and location (using store name as well as phone number|address)
capture the total
count the items (some have totals), have to teach to count.
this is where multipass of the scan comes in handy. also spellcheck the output might help.
scans needs to be saved, and also need to be uploaded somewhere. so we could rescan or teach more.
some method of verifying what ocr thinks. scan shows up with input boxes with its 'guesses'
most likely need to do the chromebook for this process. or the pi, but the pi should be doing hdmi data.