Spaces:
Building
Building
File size: 3,563 Bytes
c9f9492 7fda6bb c9f9492 f5477de d1c3219 e4aee44 08bac12 e4aee44 d1c3219 e4aee44 f5477de 08bac12 8ed001f e4aee44 d1c3219 e4aee44 8ed001f c9f9492 e4aee44 c9f9492 e4aee44 974d749 c9f9492 e4aee44 c9f9492 e4aee44 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 |
import display_gloss as dg
import synonyms_preprocess as sp
from NLP_Spacy_base_translator import NlpSpacyBaseTranslator
from flask import Flask, render_template, Response, request
from transformers import MarianMTModel, MarianTokenizer
import torch
import os
app = Flask(__name__, static_folder='static')
app.config['TITLE'] = 'Sign Language Translate'
# Set cache directory
cache_dir = "/tmp/huggingface"
if not os.path.exists(cache_dir):
os.makedirs(cache_dir, exist_ok=True)
os.environ['TRANSFORMERS_CACHE'] = cache_dir
os.environ['HF_HOME'] = cache_dir
# Force CPU usage
device = torch.device('cpu')
os.environ['CUDA_VISIBLE_DEVICES'] = ''
# Load pre-trained Korean-English translation model
model_name = "Helsinki-NLP/opus-mt-ko-en"
tokenizer = MarianTokenizer.from_pretrained(model_name, cache_dir=cache_dir)
model = MarianMTModel.from_pretrained(model_name, cache_dir=cache_dir)
model = model.to(device)
nlp, dict_docs_spacy = sp.load_spacy_values()
dataset, list_2000_tokens = dg.load_data()
def translate_korean_to_english(text):
try:
if any('\u3131' <= char <= '\u318F' or '\uAC00' <= char <= '\uD7A3' for char in text):
inputs = tokenizer(text, return_tensors="pt", padding=True)
outputs = model.generate(**inputs)
translation = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(f"Translated text: {translation}")
return translation
return text
except Exception as e:
print(f"Translation error: {e}")
return text
@app.route('/')
def index():
return render_template('index.html', title=app.config['TITLE'])
@app.route('/translate/', methods=['POST'])
def result():
if request.method == 'POST':
input_text = request.form['inputSentence']
try:
english_text = translate_korean_to_english(input_text)
if english_text == input_text and any('\u3131' <= char <= '\u318F' or '\uAC00' <= char <= '\uD7A3' for char in input_text):
raise Exception("Translation failed")
eng_to_asl_translator = NlpSpacyBaseTranslator(sentence=english_text)
generated_gloss = eng_to_asl_translator.translate_to_gloss()
gloss_list_lower = [gloss.lower() for gloss in generated_gloss.split() if gloss.isalnum()]
gloss_sentence_before_synonym = " ".join(gloss_list_lower)
gloss_list = [sp.find_synonyms(gloss, nlp, dict_docs_spacy, list_2000_tokens)
for gloss in gloss_list_lower]
gloss_sentence_after_synonym = " ".join(gloss_list)
return render_template('result.html',
title=app.config['TITLE'],
original_sentence=input_text,
english_translation=english_text,
gloss_sentence_before_synonym=gloss_sentence_before_synonym,
gloss_sentence_after_synonym=gloss_sentence_after_synonym)
except Exception as e:
print(f"Error in translation process: {e}")
return render_template('error.html', error=str(e))
@app.route('/video_feed')
def video_feed():
sentence = request.args.get('gloss_sentence_to_display', '')
gloss_list = sentence.split()
return Response(dg.generate_video(gloss_list, dataset, list_2000_tokens),
mimetype='multipart/x-mixed-replace; boundary=frame')
if __name__ == "__main__":
app.run(host="0.0.0.0", port=7860, debug=True) |