I sent robot forgeries to a handwriting expert

Comments · 9.5K

• JE

  @Jellooze · 3 years ago

  I cant believe you managed to create machine learning code for doctors handwriting on the first try

  47K

• TH

  @thelegendofme7520 · 3 years ago

  This video is the embodiment of "we do things not because they are easy, but because we thought they were"

  4.3K

• MI

  @MisaelCapri4 · 3 years ago

  What I learned from this channel over the years is that in order to do less work you have to do more work than you originally had to.

  10K

• 80

  @80HD-0 · 1 year ago

  great example of how engineers (of any type) would rather spend 10 hours automating something than an hour doing it manually.

  643

• EN

  @ennuiii · 3 years ago

  I love the "wife annoyed to be forced to help her husbands weird projects" character she pulls lmao

  5.2K

• RA

  @randomdoodles · 3 years ago

  Honestly I think that having a postcard written by a project you made is way cooler than having one hand written

  841

• IA

  @ianshook · 3 years ago

  I have to say, one of the most helpful parts of your video was when you gave up and used code off the internet. It's nice to see others realize that some other people just do things better sometimes and you don't have to re-invent the wheel every single project. Buying a plotter, borrowing code. This is how things move forward. Good luck in your new shop!

  1.2K

• GI

  @gionicol_ · 2 years ago

  This man just casually gave the best simplified explanation of how machine learning works

  285

• H2

  @H2O2FA · 3 years ago

  Ok, one major tip: natural hand writing is in fact a 3D action not just 2D, meaning that the writer exerts higher and lower pressure vertical to the paper surface as they write, which results in the pen line becoming thinner and thicker at different sections of a letter! Next try to build the Y-axis movement into that robot!!

  4.2K

• RU

  @russellinator · 3 years ago

  I'm surprised pen pressure on the paper wasn't more of a problem. Seems like the robots perfect line darkness would stand out more.

  2.8K

• ST

  @styxz5980 · 3 years ago

  just a tip when using neural networks. In the video, I noticed after every bug you fixed, the editing at least made it look like you spent ~50 hours training the RNN again. Usually, you can use smaller datasets to train the networks and see if the output is slightly acceptable before spending the 2 days training the network with the full dataset.

  3.2K

• CԼ

  @CԼIPPY · 1 year ago

  The "Garbage-O-Meter" analogy is one of the best Machine Learning analogies I have ever heard. Thanks! I will use this.

  201