Spaces:
Sleeping
Sleeping
Update ai_text_detector_valid_final.py
Browse files- ai_text_detector_valid_final.py +27 -16
ai_text_detector_valid_final.py
CHANGED
|
@@ -5,16 +5,18 @@ import requests
|
|
| 5 |
import numpy as np
|
| 6 |
|
| 7 |
# Hugging Face Token
|
| 8 |
-
HF_TOKEN = os.getenv("HF_TOKEN") #
|
|
|
|
|
|
|
| 9 |
|
| 10 |
# Headers for API
|
| 11 |
headers = {"Authorization": f"Bearer {HF_TOKEN}"} if HF_TOKEN else {}
|
| 12 |
|
| 13 |
# Multiple AI text detection models
|
| 14 |
MODELS = {
|
| 15 |
-
"DeBERTa Detector": "
|
| 16 |
-
"MonkeyDAnh":
|
| 17 |
-
"Andreas122001":
|
| 18 |
# SzegedAI handled separately since it's a Space
|
| 19 |
}
|
| 20 |
|
|
@@ -30,22 +32,31 @@ def run_hf_model(model_id, text):
|
|
| 30 |
return {"Human Probability": float(probs[0]*100), "AI Probability": float(probs[1]*100)}
|
| 31 |
except Exception as e:
|
| 32 |
return {"error": str(e)}
|
| 33 |
-
|
| 34 |
-
def run_szegedai(text):
|
| 35 |
-
"""Call the SzegedAI Space API"""
|
| 36 |
try:
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
headers=headers,
|
| 40 |
-
json={"data": [text]},
|
| 41 |
-
timeout=30
|
| 42 |
-
)
|
| 43 |
response.raise_for_status()
|
| 44 |
result = response.json()
|
| 45 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
except Exception as e:
|
| 47 |
return {"error": str(e)}
|
| 48 |
-
|
| 49 |
def detect_text(text):
|
| 50 |
results = {}
|
| 51 |
# Transformers models
|
|
@@ -53,7 +64,7 @@ def detect_text(text):
|
|
| 53 |
results[name] = run_hf_model(model_id, text)
|
| 54 |
|
| 55 |
# SzegedAI (Space)
|
| 56 |
-
results["SzegedAI Detector"] =
|
| 57 |
|
| 58 |
# Final verdict (simple rule-based)
|
| 59 |
ai_probs = []
|
|
|
|
| 5 |
import numpy as np
|
| 6 |
|
| 7 |
# Hugging Face Token
|
| 8 |
+
HF_TOKEN = os.getenv("HF_TOKEN") # Hugging Face token (optional if space is public)
|
| 9 |
+
SZEGEDAI_URL = "https://hf.space/embed/SzegedAI/AI_Detector/api/predict/"
|
| 10 |
+
HEADERS = {"Authorization": f"Bearer {HF_TOKEN}"} if HF_TOKEN else {}
|
| 11 |
|
| 12 |
# Headers for API
|
| 13 |
headers = {"Authorization": f"Bearer {HF_TOKEN}"} if HF_TOKEN else {}
|
| 14 |
|
| 15 |
# Multiple AI text detection models
|
| 16 |
MODELS = {
|
| 17 |
+
"DeBERTa Detector": "distilbert-base-uncased-finetuned-sst-2-english",
|
| 18 |
+
"MonkeyDAnh":"MonkeyDAnh/my-awesome-ai-detector-roberta-base-v4-human-vs-machine-finetune",
|
| 19 |
+
"Andreas122001":"andreas122001/roberta-academic-detector"
|
| 20 |
# SzegedAI handled separately since it's a Space
|
| 21 |
}
|
| 22 |
|
|
|
|
| 32 |
return {"Human Probability": float(probs[0]*100), "AI Probability": float(probs[1]*100)}
|
| 33 |
except Exception as e:
|
| 34 |
return {"error": str(e)}
|
| 35 |
+
def szegedai_predict(text):
|
|
|
|
|
|
|
| 36 |
try:
|
| 37 |
+
payload = {"data": [text]}
|
| 38 |
+
response = requests.post(SZEGEDAI_URL, json=payload, headers=HEADERS, timeout=30)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
response.raise_for_status()
|
| 40 |
result = response.json()
|
| 41 |
+
|
| 42 |
+
raw = result["data"][0] # e.g. "Human Probability: 99.83% | AI Probability: 0.17%"
|
| 43 |
+
|
| 44 |
+
human_match = re.search(r"Human[^0-9]*([\d.]+)%", raw)
|
| 45 |
+
ai_match = re.search(r"AI[^0-9]*([\d.]+)%", raw)
|
| 46 |
+
|
| 47 |
+
if human_match and ai_match:
|
| 48 |
+
human_prob = float(human_match.group(1))
|
| 49 |
+
ai_prob = float(ai_match.group(1))
|
| 50 |
+
return {
|
| 51 |
+
"Human Probability": round(human_prob, 2),
|
| 52 |
+
"AI Probability": round(ai_prob, 2),
|
| 53 |
+
}
|
| 54 |
+
else:
|
| 55 |
+
return {"error": f"Unexpected response: {raw}"}
|
| 56 |
+
|
| 57 |
except Exception as e:
|
| 58 |
return {"error": str(e)}
|
| 59 |
+
|
| 60 |
def detect_text(text):
|
| 61 |
results = {}
|
| 62 |
# Transformers models
|
|
|
|
| 64 |
results[name] = run_hf_model(model_id, text)
|
| 65 |
|
| 66 |
# SzegedAI (Space)
|
| 67 |
+
results["SzegedAI Detector"] = szegedai_predict(text)
|
| 68 |
|
| 69 |
# Final verdict (simple rule-based)
|
| 70 |
ai_probs = []
|