Looks promising—thanks for sharing! Is it possible to include non-Python binaries somehow? Similar to how it is done with PyInstaller with the --add-binary argument?
P.S.: It has already been mentioned, but I also find the "cowsay" example more confusing than informative. "cowsay" isn't even written in Python, but in Perl, isn't it? Maybe a simple hello-world.py example would work better :)
As others mentioned, Tesseract is SOTA in FOSS OCR. It also still is being developed, improving slow but constantly.
The main issue for a use-case like NormCap are the trained models: they are optimized for images of _printed_ text and layouts, which is different from on-screen-text in many aspects. Unfortunately, I don't have the resources to train my own models.
Cuneiform was a long time competitor, but afaik development there is stalled.
I agree, there are way better cloud based and proprietary OCR solutions out there. But Tesseract still seems to deliver the best results among the FOSS tools, doesn't it?
Back in the days, Cuneiform got close to Tesseract's performance, but AFAIK it wasn't developed further...
Does anyone else know other promising open-source OCR engines?
I didn't know about the story yet, but the reality was already very close to that terrible scenario: https://en.wikipedia.org/wiki/Bachem_Ba_349
(One of the launch sites is quite close to where I live)
Would love to see statistics from the other perspective: Which animals are killed most by humans? Besides insects, I would guess fish and chicken are quite high...
I see a lot of statements like this (also here in the comments), but is there actual any evidence behind those? Any research results, which has been replicated?
If yes, it would be nice to see it together with such statements, to back them, because I have the feeling most of this is just anecdotical or ideas introduced by some kind of start-up gurus...
Please be careful not to generalize when proposing solutions, because their is not the depression(1): Depressions can be caused by very different things, can cause somewhat different symptoms and certainly can be solved using very different approaches. I met a lot of of depressed people, most them got out of them with different approaches (meds, therapy, changes in their environment).
I myself tried the approach mentioned, during therapy, without any success: Achieving something just didn't produce any positive effect for me. Then I got treated (as last resort) with ECT, my condition improved temporarily, and suddenly, within the same therapy, I started to feel an effect and was able to stabilize.
In short: Propsing "Do <x>, it most likely will help you" is not a good idea, as it is quite depressing for the depressed (at least for me it was). Maybe it's better to say sth like "Try <x>, maybe it helps, if not, there is also <y> or <z>. There is a lot that can be tried, maybe something might help you. "
(1) A doctor explained it to me like that: "depression" is more like a term to group together conditions that partially share symptoms, and that is it basically used for handling the insurance stuff etc. For treatment (at least for severe depressions), the term is too rough to be useful. E.g. a lot of severely depressed people also show symptoms of psychosis as side effect.
EDIT: Formatting and reformulated a bit (I think I was too harsh)
Might this be similar to the stress/shock introduced partial memory loss, as it often happens with heavy accidents, where the accident itself isn't remembered by the patient?
Thank you for summarising and thx to OP for publishing his work. The combination of both was a nice opportunity for me (and others?) to learn something from mistakes. :-)
An unknown chessplayer bet a lot of money, that he could play live & simultaniously against two grandmasters and achieve at least 1 point. The two matches were played in different rooms next to each other. No one except the unknown player was allowed to switch rooms during the matches (to not disturb the grandmasters). Obviously he played White against one grandmaster, Black against the other, and won his bet.
As I sometimes tell this anecdote, I wonder what its actual source might be?
I'm pretty confident, this never happend in real, as I expect no grandmaster could get scammed that easily. But the story could be rooted in a "catch me if you can"-like movie or novel or something.
If someone knows more, I would be happy for a hint...
In my circle of friends, there was a similar tragic case: Someone suicidal took deadly poison with delayed effect (5-10min), immediately regretted it and called one of my friends for help. (Sadly, the ambulance couldn't save him.)
The very pessimistic philosopher Emil Cioran (author of e.g. De l'inconvénient d'être né, "The Trouble With Being Born") who beside his suicidal thoughts died at age 84 once said in an interview, that suicidal thoughts are paradoxically the reason, why he stayed alive: The knowledge of the choice to end existence if the suffering is too much makes it possible to endure its insufferableness.
Maybe it is this choice, to live or not, that makes a lot things bearable? And it's much harder if this choice is gone for whatever reason, in one direction or another?
I'm working in a Data Science Team and like my job a lot. There are some frustrating aspects, so, most of them rooted in the hype around ML / AI. Just some examples:
1. The massive hype around ML has produced unrealistic expectation.
We often face the issue, that customers are unhappy with our results, because their high expectations: "Uh, your model has only 78% accuracy, can't you do better?" (No, not without adequate data and resources!) You basically disappoint people very often.
2. There are lot's of fraudsters in the game, that might get the fame.
I have seen data scientists being applauded, because they claimed to get "99,7% accuracy" for an regression problem. How the hell did they calculate that? Accuracy isn't even a good metric for regression problems! Of course those models usually don't stand reality, but that doesn't seem to be relevant...
3. The work might not have a relevant impact.
Often, we do prototypes to tackle problems, that don't even have enough impact to ever become profitable. We are set on it, because some manager has been told to "Leverage AI to improve Business". As a consequence, when it becomes obvious, that the resources needed to run something in production will never create a positive ROI, the project remains a prototype. Of course our Team knows & communicates that often from the beginning, but it doesn't even seem to matter, as long as anyone can put "working on AI" on some PowerPoint slide.
4. Most of the work is "boring" data preparation.
The "cool" modelling part of our work, where you design architectures and evaluate algorithms usually is ~5% of our implementation time. Most of the time is spent in preparing the data to be suitable input for the model. (I myself actually like data prep as a part of the process, but I know lots of colleagues don't).
This is my experience from working in a very large but not digital native company. I'd expect it to be different at Amazon, Netflix, Google etc. And I'd be interested to hear if data scientist from other companies face similar or different issues...
It even could work offline, cache/store data and send it when back online.