MNIST start dates
I am currently working on a handwriting recognition problem. I found that many standard-based algorithms apply some preprocessing techniques to the timestamp date such as correction and jitter (I don't know what "jitter" is). I've googled for a long time but didn't find anything about specific methods or algorithms that could be used for datetime mnist. Anyone have any ideas on how to solve this problem? Many thanks!
source to share
I've Googled it for a long time and is hard to find this, but never succeeded until I saw this article:
Teow, Loo-Nin and Kia-Fock Loe. "Robust vision-based features and grading scheme for offline handwritten digit recognition." Pattern Recognition 35.11 (2002): 2355-2364.
Read section 4.2. They described a procedure for deslanting their images. I'm not sure if this is the same as deskewing , but it is the closest thing I could find.
Hope you enjoy this help if anyone can confirm that deslanting is the same as deskewing .
source to share
Written letter taken from LeCun 1998 article [ http://yann.lecun.com/exdb/publis/pdf/lecun-98.pdf] .
The lattice calculates the second moment of inertia of the pixels (assuming the foreground pixel as one and the background pixel as zero) and shifts the image horizontally, shifting the lines so that the main axis is vertical.
source to share