Thank you so much! I got it working. Didn't think about inverting the
images.
Best,
Cailey
On Sunday, November 1, 2020 at 11:59:00 AM UTC-5 Cailey McVay wrote:
> How did you invert the image? And is there a code I can use to invert the
> rest of my images to try with more sampl
- --psm 6
> 063.433
> $ tesseract legacy-300.jpg - --psm 6
> 063.433
> $ tesseract legacy-144.jpg - --psm 6
> 063.433
>
>
>
> On Sun, Nov 1, 2020 at 8:37 PM Cailey McVay
> wrote:
>
>> Here is an example of the sample image. I believe we are using the lega
the time.
>
> You haven't shared a sample image. Sometimes preprocessing the images,
> using a whitelist in case of limited character set can be the solution
> rather than training.
>
> On Sun, Nov 1, 2020, 03:29 Cailey McVay wrote:
>
>> Hello!
>> I am working on
Hello!
I am working on a project that is trying to read borehole video depths. We
trained a new language to read these numbers called NTS. When we use
tesseract on the images without the trained language we receive outputs
that are accurate about 50% of the time. However when we use the new
Hello!
I am working on a project that is trying to read borehole video depths. We
trained a new language to read these numbers called NTS. When we use
tesseract on the images without the trained language we receive outputs
that are accurate about 50% of the time. However when we use the new
Hello,
I am trying to fine-tune Tesseract with a new language to interpret these
images posted below. I am able to create the box tr file with:
tesseract NTS.tif NTS --psm6 nobatch box.train.
And the unicharset file with:
unicharset_extractor NTS.box.
However when I get to the command for
6 matches
Mail list logo
