URGENT DEADLINE: NEED HELP WITH NEW LANGUAGE, PLEASE RESPOND

Cailey McVay

unread,

Oct 31, 2020, 5:59:28 PM10/31/20

to tesseract-ocr

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 language, we receive no output at all. Is it possible that we overtrained tesseract to not recognize any of the images? I will attach below our box file, unicharset file, box trained file, pffmtable file, and normproto file. Our shapetable file processes but then returns an empty file. Could something be wrong with our shapetable? And if so, how could we fix that?

Box File for the first five images:

0 3 1 14 19 0

9 18 0 29 20 0

3 33 1 46 19 0

. 50 1 56 19 0

2 64 1 75 19 0

5 76 1 93 19 0

2 92 1 111 19 0

0 4 1 15 19 1

8 19 1 30 19 1

3 34 1 46 19 1

. 54 1 57 5 1

4 65 1 77 19 1

1 82 1 91 19 1

4 96 1 107 19 1

0 3 1 15 19 2

8 19 1 30 19 2

6 34 1 46 19 2

. 53 1 57 5 2

8 65 1 77 19 2

3 80 1 91 19 2

9 95 1 107 19 2

0 4 1 15 19 3

8 17 1 31 19 3

8 32 1 46 19 3

. 52 2 58 8 3

1 64 0 77 20 3

8 80 1 91 19 3

5 96 1 107 19 3

0 3 1 15 19 4

8 19 1 30 19 4

7 34 1 47 19 4

. 53 1 58 9 4

5 65 1 77 19 4

6 80 1 92 19 4

4 95 0 109 20 4

0 4 1 15 19 5

7 19 1 30 19 5

5 34 1 46 19 5

. 53 1 57 5 5

3 65 1 76 19 5

1 82 1 90 19 5

3 96 1 107 19 5

Unicharset:

14

NULL 0 Common 0

Joined 7 0,255,0,255,0,0,0,0,0,0 Latin 1 0 1 Joined # Joined [4a 6f 69 6e 65 64 ]a

0 8 0,255,0,255,0,0,0,0,0,0 Common 3 2 3 0 # 0 [30 ]0

9 8 0,255,0,255,0,0,0,0,0,0 Common 4 2 4 9 # 9 [39 ]0

3 8 0,255,0,255,0,0,0,0,0,0 Common 5 2 5 3 # 3 [33 ]0

. 22 0,255,0,255,0,0,0,0,0,0 Common 6 6 6 . # . [2e ]p

2 8 0,255,0,255,0,0,0,0,0,0 Common 7 2 7 2 # 2 [32 ]0

5 8 0,255,0,255,0,0,0,0,0,0 Common 8 2 8 5 # 5 [35 ]0

8 8 0,255,0,255,0,0,0,0,0,0 Common 9 2 9 8 # 8 [38 ]0

4 8 0,255,0,255,0,0,0,0,0,0 Common 10 2 10 4 # 4 [34 ]0

1 8 0,255,0,255,0,0,0,0,0,0 Common 11 2 11 1 # 1 [31 ]0

6 8 0,255,0,255,0,0,0,0,0,0 Common 12 2 12 6 # 6 [36 ]0

7 8 0,255,0,255,0,0,0,0,0,0 Common 13 2 13 7 # 7 [37 ]0

NTS.font.exp0.tr file:

font 0 3 1 14 19 0

4

mf 16

-0.085041896 0.30783021 0.27617577 0 0 0

-0.25234067 0.27376649 0.089746617 0.13718249 0 0

-0.28155157 0.0045010448 0.47040343 0.25 0 0

-0.25234067 -0.26476437 0.08974655 0.36281759 0 0

-0.085041896 -0.29882804 0.27617577 0.5 0 0

-0.031931162 -0.21447986 0.1730229 0.96998096 0 0

-0.11690831 0.020721853 0.43796182 0.75 0 0

-0.031931162 0.23970276 0.1699543 0.5 0 0

0.24424461 0.072628468 0.47339222 0.76789355 0 0

0.1353676 0.30783021 0.16464323 0 0 0

0.10615671 0.18941826 0.14627755 0.37934926 0 0

0.15926743 -0.011719763 0.30170703 0.25 0 0

0.10615671 -0.19663697 0.12619166 0.090763755 0 0

0.1353676 -0.29882804 0.16464323 0.5 0 0

0.27079996 -0.26476437 0.12619169 0.59076369 0 0

0.29735535 -0.19663697 0.086383387 0.85538673 0 0

cn 1

0.36328125 0.35781249 0.2421875 0.1484375

if 73

133 69 248

119 72 248

104 75 248

97 82 192

97 95 192

97 107 192

97 120 192

97 132 192

97 145 192

97 157 192

97 170 192

97 182 192

104 188 128

119 188 128

133 188 128

135 206 0

123 206 0

111 206 0

99 206 0

88 206 0

76 206 0

66 201 35

59 193 35

55 182 64

55 168 64

55 155 64

55 142 64

55 128 64

55 115 64

55 101 64

55 88 64

55 75 64

59 64 93

66 55 93

76 51 128

88 51 128

99 51 128

111 51 128

123 51 128

135 51 128

145 184 97

154 175 97

163 167 97

168 156 64

168 143 64

168 130 64

168 118 64

168 105 64

168 92 64

163 82 23

154 77 23

145 71 23

148 51 128

162 51 128

176 51 128

187 53 151

196 59 151

205 65 151

207 72 219

200 81 219

196 92 192

196 105 192

196 118 192

196 130 192

196 143 192

196 156 192

195 168 204

191 179 204

188 190 204

184 200 204

176 206 0

162 206 0

148 206 0

tb 1

64 251 114

pffmtable:

NULL 0

Joined 0

|Broken|0|1 0

0 0

9 0

3 0

. 0

2 0

5 0

8 0

4 0

1 0

6 0

7 0

NTS.normproto file:

linear essential -0.250000 0.750000

linear non-essential 0.000000 1.000000

linear essential 0.000000 1.000000

0 1

significant elliptical 34

0.364775 0.371404 0.241039 0.150391

0.000400 0.000416 0.000400 0.000400

9 1

significant elliptical 13

0.372897 0.418750 0.241286 0.157752

0.000400 0.004734 0.000400 0.001087

3 1

significant elliptical 16

0.365479 0.385596 0.247070 0.143799

0.000400 0.003148 0.000400 0.000702

. 1

significant elliptical 27

0.081019 0.055483 0.060619 0.050492

0.000400 0.000400 0.000400 0.000400

2 1

significant elliptical 10

0.354297 0.359492 0.248828 0.138672

0.000400 0.000400 0.000400 0.000400

5 1

significant elliptical 10

0.363672 0.350859 0.248047 0.144922

0.000400 0.000400 0.000400 0.000400

8 1

significant elliptical 19

0.365543 0.378536 0.234786 0.141653

0.000400 0.000400 0.000400 0.000400

4 1

significant elliptical 9

0.325521 0.274219 0.215278 0.128038

0.000400 0.000400 0.000400 0.000400

1 1

significant elliptical 11

0.320312 0.217259 0.248580 0.091974

0.000400 0.000400 0.000400 0.000400

6 1

significant elliptical 20

0.360156 0.370703 0.238281 0.143164

0.000400 0.000400 0.000400 0.000400

7 1

significant elliptical 20

0.448633 0.243359 0.242969 0.113477

0.000400 0.000400 0.000400 0.000400

Shree Devi Kumar

unread,

Oct 31, 2020, 11:13:41 PM10/31/20

to tesseract-ocr

Are you trying to train for the legacy tesseract engine?

--
You received this message because you are subscribed to the Google Groups "tesseract-ocr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tesseract-oc...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/9e3a6851-0311-4148-af1f-b61999f38977n%40googlegroups.com.

Shree Devi Kumar

unread,

Oct 31, 2020, 11:15:46 PM10/31/20

to tesseract-ocr

>When we use tesseract on the images without the trained language we receive outputs that are accurate about 50% of 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 <cailey.m...@dartmouth.edu> wrote:

--

Cailey McVay

unread,

Nov 1, 2020, 10:07:51 AM11/1/20

to tesseract-ocr

Here is an example of the sample image. I believe we are using the legacy engine. Does this help?

Screen Shot 2020-11-01 at 10.05.07 AM.png

Shree Devi Kumar

unread,

Nov 1, 2020, 10:55:00 AM11/1/20

to tesseract-ocr

Invert the image. Results using tessdata_best/eng - LSTM engine

$ tesseract legacy-invert.jpg - --psm 6
063.433
$ tesseract legacy-300.jpg - --psm 6
063.433
$ tesseract legacy-144.jpg - --psm 6
063.433

To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/c09d4786-595e-4e49-b5c6-b7ded4bee47fn%40googlegroups.com.

--

____________________________________________________________
भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com

legacy-144.jpg

legacy-300.jpg

legacy-invert.jpg

Cailey McVay

unread,

Nov 1, 2020, 11:59:00 AM11/1/20

to tesseract-ocr

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 sample data?

Cailey McVay

unread,

Nov 3, 2020, 11:56:42 AM11/3/20

to tesseract-ocr

Thank you so much! I got it working. Didn't think about inverting the images.

Best,

Cailey

Reply all

Reply to author

Forward