Spaces:
Running
Running
Update appImage.py
Browse files- appImage.py +51 -23
appImage.py
CHANGED
@@ -1,32 +1,60 @@
|
|
1 |
-
|
2 |
-
from
|
3 |
-
import
|
|
|
4 |
from PIL import Image
|
|
|
5 |
from gtts import gTTS
|
|
|
|
|
|
|
|
|
|
|
6 |
|
7 |
-
|
8 |
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
image_path = tmp.name
|
14 |
|
15 |
-
|
16 |
-
|
|
|
|
|
|
|
|
|
|
|
17 |
|
18 |
-
|
|
|
|
|
19 |
|
20 |
-
|
21 |
-
if audio_path:
|
22 |
-
result["audioUrl"] = f"/files/{os.path.basename(audio_path)}"
|
23 |
-
return result
|
24 |
|
25 |
-
def text_to_speech(text: str):
|
26 |
try:
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
from fastapi import FastAPI
|
3 |
+
from fastapi.responses import RedirectResponse, JSONResponse, FileResponse
|
4 |
+
import os
|
5 |
from PIL import Image
|
6 |
+
from transformers import ViltProcessor, ViltForQuestionAnswering, pipeline
|
7 |
from gtts import gTTS
|
8 |
+
import easyocr
|
9 |
+
import torch
|
10 |
+
import tempfile
|
11 |
+
import numpy as np
|
12 |
+
from io import BytesIO
|
13 |
|
14 |
+
app = FastAPI()
|
15 |
|
16 |
+
vqa_processor = ViltProcessor.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
|
17 |
+
vqa_model = ViltForQuestionAnswering.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
|
18 |
+
captioner = pipeline("image-to-text", model="nlpconnect/vit-gpt2-image-captioning")
|
19 |
+
reader = easyocr.Reader(['en', 'fr'])
|
|
|
20 |
|
21 |
+
def classify_question(question: str):
|
22 |
+
q = question.lower()
|
23 |
+
if any(w in q for w in ["text", "say", "written", "read"]):
|
24 |
+
return "ocr"
|
25 |
+
if any(w in q for w in ["caption", "describe", "what is in the image"]):
|
26 |
+
return "caption"
|
27 |
+
return "vqa"
|
28 |
|
29 |
+
def answer_question_from_image(image, question):
|
30 |
+
if image is None or not question.strip():
|
31 |
+
return "Please upload an image and ask a question.", None
|
32 |
|
33 |
+
mode = classify_question(question)
|
|
|
|
|
|
|
34 |
|
|
|
35 |
try:
|
36 |
+
if mode == "ocr":
|
37 |
+
result = reader.readtext(np.array(image))
|
38 |
+
answer = " ".join([entry[1] for entry in result]) or "No readable text found."
|
39 |
+
|
40 |
+
elif mode == "caption":
|
41 |
+
answer = captioner(image)[0]['generated_text']
|
42 |
+
|
43 |
+
else:
|
44 |
+
inputs = vqa_processor(image, question, return_tensors="pt")
|
45 |
+
with torch.no_grad():
|
46 |
+
outputs = vqa_model(**inputs)
|
47 |
+
predicted_id = outputs.logits.argmax(-1).item()
|
48 |
+
answer = vqa_model.config.id2label[predicted_id]
|
49 |
+
|
50 |
+
tts = gTTS(text=answer)
|
51 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp:
|
52 |
+
tts.save(tmp.name)
|
53 |
+
return answer, tmp.name
|
54 |
+
|
55 |
+
except Exception as e:
|
56 |
+
return f"Error: {e}", None
|
57 |
+
|
58 |
+
@app.get("/")
|
59 |
+
def home():
|
60 |
+
return RedirectResponse(url="/templates/home.html")
|