sign-language-processing/spoken-to-signed-translation

New another language to add or change.

Opened this issue · 7 comments

GLWmax commented

Error:
download_lexicon.py --name 'vgt' --directory ./vgt
usage: download_lexicon.py [-h] --name {signsuisse} --directory DIRECTORY
download_lexicon.py: error: argument --name: invalid choice: 'vgt' (choose from 'signsuisse')

I have add 'VGT'

from .types import Gloss
from .common import load_spacy_model

LANGUAGE_MODELS_SPACY = {
"de": "de_core_news_lg",
"fr": "fr_core_news_lg",
"vgt": "vgt_core_news_lg",
"en": "en_core_web_lg",
}

def text_to_gloss(text: str, language: str, ignore_punctuation: bool = False) -> Gloss:

if language not in LANGUAGE_MODELS_SPACY:
    raise NotImplementedError("Don't know language '%s'." % language)

model_name = LANGUAGE_MODELS_SPACY[language]

# disable unnecessary components to make lemmatization faster

spacy_model = load_spacy_model(model_name, disable=("parser", "ner"))

doc = spacy_model(text)

glosses = []  # type: Gloss

for token in doc:
    if ignore_punctuation is True:
        if token.is_punct:
            continue

    gloss = (token.text, token.lemma_)
    glosses.append(gloss)

return glosses
AmitMY commented

I see that you are trying to use this repository with VGT.

The download_lexicon script does not support any VGT dataset, so to support Flemish, you would have to go through the following process:

  1. Collect a lexicon (Download videos from https://vlaamsegebarentaal.be/signbank/signs/show_all/ or collect your own)
  2. Extract poses using this library and the command video_to_pose --format mediapipe -i example.mp4 -o example.pose
  3. Construct a lexicon CSV file with the words, matching the poses, for example https://github.com/sign-language-processing/spoken-to-signed-translation/blob/main/assets/dummy_lexicon/index.csv
path,spoken_language,signed_language,start,end,words,glosses,priority
sgg/kleine.pose,de,sgg,0,0,kleine,Kleine,0
sgg/kinder.pose,de,sgg,0,0,kinder,Kinder,0

Now, once you have this index.csv, under a directory called, let's say, lexicon, you can run for example:

text_to_gloss_to_pose \
  --text "Hallo mijn naam is john." \
  --glosser "simple" \
  --lexicon "lexicon" \
  --spoken-language "nl" \
  --signed-language "vgt" \
  --pose "quick_test.pose"
KhayitboevElbekjon commented

hello,I have one problem, look at.

download_lexicon
--name
--directory <path_to_directory>

What should I put in "name" and "directory" in this code?

KhayitboevElbekjon commented

which file should i run to use this program?

AmitMY commented

hello,I have one problem, look at.

download_lexicon --name --directory <path_to_directory>

What should I put in "name" and "directory" in this code?

the only dataset available in this repository is signsuisse.
If you have a further issue that is not related to the issue at hand, please create a different issue.

KhayitboevElbekjon commented