hchcsuim commited on
Commit
65e0c0f
·
verified ·
1 Parent(s): 70b006c

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +47 -19
app.py CHANGED
@@ -12,35 +12,63 @@ def predict(input_img):
12
  predictions = pipe(input_img)
13
  return input_img, {p["label"]: p["score"] for p in predictions}
14
 
15
- # 模型說明與免責聲明(會放在最下方)
16
  model_card = """
17
  ---
18
 
19
- ## 🧠 Model Description / 模型簡介
 
 
20
 
21
- This model is a fine-tuned version of `microsoft/swin-tiny-patch4-window7-224` on the imagefolder dataset.
22
- It classifies face images into two categories: **AI-generated** and **Not AI-generated**.
23
 
24
- 本模型是以 `microsoft/swin-tiny-patch4-window7-224` 微調訓練而成,可將臉部圖片分類為「AI生成」或「非AI生成」。
25
-
26
- 🔗 [View on Hugging Face 模型主頁](https://huggingface.co/hchcsuim/FaceAIorNot)
27
 
28
  ### 📊 Evaluation Results / 模型評估
29
 
30
- - Accuracy 準確率: **99.35%**
31
- - Precision 精確率: **99.25%**
32
- - Recall 召回率: **99.47%**
33
- - F1-score: **99.36%**
34
- - Loss: **0.0233**
 
 
 
 
 
 
 
 
 
35
 
36
- 訓練資料包含 105,330 張臉部圖片(50% 為真實臉部,50% AI 合成臉),來自 17 個資料集、14 AI 生成技術,90% 作為訓練集、10% 為測試集。
 
 
 
37
 
38
- ### ⚠️ Disclaimer / 免責聲明
 
 
 
39
 
40
- This model is for research and educational use.
41
- It may not be 100% accurate, and **should not be used for critical decisions, such as identity verification, legal judgement, or public shaming**.
 
 
 
 
 
 
 
 
 
 
 
 
 
42
 
43
- 本模型僅供研究與教育用途參考,**結果非百分之百準確,請勿用於身分驗證、法律判斷、公開揭露等重要用途**。
 
44
  """
45
 
46
  # Gradio 介面設定(雙語版)
@@ -51,10 +79,10 @@ gradio_app = gr.Interface(
51
  gr.Image(label="🖼️ Input Image / 輸入圖片"),
52
  gr.Label(label="🔍 Classification Result / 判斷結果", num_top_classes=2)
53
  ],
54
- title="FaceAIorNot | 真人臉,還是 AI 生成臉?",
55
  description=(
56
  "🤖 Upload or take a face photo to see if it's AI-generated or real.\n"
57
- "🧑 上傳或拍攝一張臉部照片,判斷是真人還是 AI 合成圖。"
58
  ),
59
  article=model_card,
60
  allow_flagging="never"
 
12
  predictions = pipe(input_img)
13
  return input_img, {p["label"]: p["score"] for p in predictions}
14
 
15
+ # 精簡中英文對照版說明
16
  model_card = """
17
  ---
18
 
19
+ ## 🧠 Model Description / 模型簡介
20
+ This model classifies face images into two categories: **AI-generated** and **Not AI-generated**.
21
+ 本模型將臉部圖片分類為「AI生成」或「非AI生成」。
22
 
23
+ 🔗 Model Homepage / 模型主頁:https://huggingface.co/hchcsuim/FaceAIorNot
 
24
 
25
+ ---
 
 
26
 
27
  ### 📊 Evaluation Results / 模型評估
28
 
29
+ - Accuracy 準確率: **99.35%**
30
+ - Precision 精確率: **99.25%**
31
+ - Recall 召回率: **99.47%**
32
+ - F1-score F1 分數: **99.36%**
33
+ - Loss 損失: **0.0233**
34
+
35
+ Dataset: 105,330 face images (50% real / 50% AI), from 17 datasets, 14 generation techniques, 90% train / 10% test
36
+ 資料集共 105,330 張臉部圖片(50% 真人 / 50% AI),來自 17 個資料集、14 種生成技術,90% 訓練 / 10% 測試
37
+
38
+ ---
39
+
40
+ ### ⚠️ Disclaimer / 免責聲明
41
+ This model is for research and educational use only. Not 100% accurate.
42
+ 請僅用於研究與教育用途,結果非百分之百準確。
43
 
44
+ Do not use for identity verification, legal judgment, or public exposure.
45
+ 請勿用於身分驗證、法律判斷或公開揭露等用途。
46
+
47
+ ---
48
 
49
+ ### 👨‍💻 About the Developer / 關於開發者
50
+ **Hung Chih Hsiang(洪誌翔)**
51
+ Master’s student in Information Management, Cheng Shiu University.
52
+ 正修科技大學 資訊管理所碩士生。
53
 
54
+ Focus: deepfake detection, snake recognition, system development, DevOps
55
+ 專長:深度偽造辨識、蛇類辨識、系統開發與運維
56
+
57
+ ---
58
+
59
+ ### 🤝 Open to Collaborations / 歡迎合作
60
+
61
+ I'm open to:
62
+ 我歡迎以下方向的合作:
63
+
64
+ - Research or commercial AI projects / AI 研究或商業應用
65
+ - Model training, optimization, deployment / 模型訓練、優化與部署
66
+ - System development, MLOps & DevOps / 系統開發與 MLOps、DevOps 整合
67
+
68
+ ---
69
 
70
+ 📧 Email: [email protected]
71
+ 🐙 GitHub: https://github.com/hchcsuim
72
  """
73
 
74
  # Gradio 介面設定(雙語版)
 
79
  gr.Image(label="🖼️ Input Image / 輸入圖片"),
80
  gr.Label(label="🔍 Classification Result / 判斷結果", num_top_classes=2)
81
  ],
82
+ title="FaceAIorNot | 真人臉,還是 AI 生成人臉?",
83
  description=(
84
  "🤖 Upload or take a face photo to see if it's AI-generated or real.\n"
85
+ "🧑 上傳或拍攝一張臉部照片,判斷是真人還是 AI 生成圖。"
86
  ),
87
  article=model_card,
88
  allow_flagging="never"