-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathDatasets_kor.html
More file actions
280 lines (242 loc) · 22.6 KB
/
Datasets_kor.html
File metadata and controls
280 lines (242 loc) · 22.6 KB
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
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>DASH LAB - Datasets</title>
<!-- External Libraries -->
<script src="https://cdn.tailwindcss.com"></script>
<link href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.0.0/css/all.min.css" rel="stylesheet">
<!-- Custom Shared Resources -->
<link rel="stylesheet" href="css/style.css">
<script src="js/common.js"></script>
</head>
<body class="bg-gray-50 flex flex-col min-h-screen">
<div class="container mx-auto px-4 pt-8 max-w-6xl flex-grow">
<!-- Language Link -->
<p class="text-right mb-5">
<a href="Datasets" class="text-blue-700 font-bold hover:underline">English →</a>
</p>
<!-- COCO Spliced Datasets -->
<div class="section">
<div class="dash-card-dataset">
<h3 class="dash-header-dataset">📦 COCO Spliced Datasets</h3>
<p class="leading-relaxed text-gray-700 mb-4">우리는 <a href="https://cocodataset.org/#home" target="_blank" class="text-blue-600 hover:text-blue-800">COCO dataset</a> to generate 데이터셋을 활용하여 조작된 데이터셋을 생성했습니다. 이 데이터셋에서 제공된 레이블 (정답 마스크)이 포함되어 있어, 먼저 마스크를 적용해 원본 이미지에서 원하는 부분을 식별했습니다. 그런 다음 이 특정 영역을 다른 이미지에 적용하여 조작했습니다. 각 이미지에는 약 8~10개의 조작용 객체가 사용되었으며, 그 결과 약 90만 개의 조작된 이미지가 생성되었습니다.</p>
<p class="text-center mt-5">
<img loading="lazy" src="" onerror="this.src=getImg('/img/Publications/Screen Shot 2023-12-11 at 2.15.08 PM.png')" alt="COCO Spliced Example" class="max-w-full h-auto rounded-lg shadow-lg mx-auto" />
</p>
</div>
</div>
<!-- Satellite Forgery Image Dataset -->
<div class="section">
<div class="dash-card-dataset">
<h3 class="dash-header-dataset">🛰️ Satellite Forgery Image Dataset</h3>
<p class="leading-relaxed text-gray-700 mb-4">우리는<a href="http://deepglobe.org/" target="_blank" class="text-blue-600 hover:text-blue-800">DeepGlop dataset</a> 데이터셋과<a href="https://openaccess.thecvf.com/content_CVPRW_2020/papers/w39/Horvath_Manipulation_Detection_in_Satellite_Images_Using_Deep_Belief_Networks_CVPRW_2020_paper.pdf" target="_blank" class="text-blue-600 hover:text-blue-800">Deep Belief networks</a>. 네트워크에서 제안된 방법을 사용하여 위성 조작 이미지를 생성했습니다. 해상도가 1000×1000인 293개의 정사보정(orthorectified) 이미지를 수집하였으며, 이 중 100개의 이미지를 사용해 조작된 이미지를 만들었습니다. 19개의 다양한 객체를 100개의 이미지에 합성하여 총 500개의 조작된 이미지와 정답 마스크를 생성했습니다. 19개의 객체에는 로켓, 비행기, 드론 이미지 등이 포함되어 있습니다. 아래 그림은 조작된 데이터셋의 예시를 보여줍니다.</p>
<p class="text-center mt-5">
<img loading="lazy" src="" onerror="this.src=getImg('/img/Publications/satellite_forgery.png')" alt="Satellite Forgery Example" class="max-w-full h-auto rounded-lg shadow-lg mx-auto" />
</p>
</div>
</div>
<!-- RWDF-23 Dataset -->
<div class="section">
<div class="dash-card-dataset">
<h3 class="dash-header-dataset">🎬 RWDF-23 Dataset</h3>
<p class="leading-relaxed text-gray-700 mb-4">RWDF-23 데이터셋은 21개국에서 4가지 언어를 대상으로 하는 2,000개의 딥페이크 비디오로부터 수집되었고, 수집된 플랫폼은 Reddit, YouTube, TikTok 그리고 Bilibili입니다. 저희는 기존 연구를 넘어 데이터셋의 스코프를 확장함으로써, 광범위한 real-world 딥페이크 콘텐츠를 캡처하였고, 이는 끊임없이 진화하는 온라인 플랫폼 환경을 반영합니다. 또한, creator, manipulation strategy, purpose, real-world content production method 등 다양한 측면을 포괄하는 종합적인 분석을 진행하였습니다. 이를 통해, 다양한 맥락에서 딥페이크의 뉘앙스와 특성에 대한 통찰력을 얻을 수 있을 것입니다. 마지막으로, 비디오 콘텐츠 외에도 시청자의 댓글과 상호작용을 수집하여, 딥페이크 콘텐츠에 대한 사용자들의 참여를 탐색할 수 있게 하였습니다. </p>
<div class="mt-5 p-4 bg-blue-100 rounded-lg border-l-4 border-blue-600 text-center">
<strong class="text-gray-800">📝 To obtain the dataset, please fill out the form <a href="https://docs.google.com/forms/d/e/1FAIpQLScsxskSEI0LkmUdI7ClAqs-xslyviDNoKHhiZC3FsBqFG4NJA/viewform" target="_blank" class="text-blue-700 hover:text-blue-900 underline">HERE</a></strong>
</div>
<p class="text-center mt-5">
<img loading="lazy" src="" onerror="this.src=getImg('/img/Publications/rwdf23_cikm23.png')" alt="RWDF-23 Example" class="max-w-full h-auto rounded-lg shadow-lg mx-auto" />
</p>
</div>
</div>
<!-- FakeAVCeleb Dataset -->
<div class="section">
<div class="dash-card-dataset">
<h3 class="dash-header-dataset">🎭 FakeAVCeleb Dataset</h3>
<p class="leading-relaxed text-gray-700 mb-4">FakeAVCeleb에서, 저희는 딥페이크 비디오뿐만 아니라 립싱크로부터 합성된 fake 오디오도 포함하는 새로운 Audio-Video 딥페이크 데이터셋을 제안하였습니다. 저희의 FakeAVCeleb은 가장 유명한 최신 딥페이크 제작 방식으로 만들어졌습니다. 더 사실적인 데이터셋을 만들기 위해, 저희는 4개 인종(Caucasian, Black, East Asian, South Asian) 연예인의 YouTube 영상을 선택하여 인종에 bias 되는 문제를 해결하였습니다.</p>
<p class="text-center mt-5">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/FakeAVCeleb/fakeceleb_nips2021.png')" alt="FakeAVCeleb Example" class="max-w-full h-auto rounded-lg shadow-lg mx-auto" />
</p>
</div>
</div>
<!-- VFP290K Dataset -->
<div class="section">
<div class="dash-card-dataset">
<h3 class="dash-header-dataset">🚨 VFP290K Dataset</h3>
<p class="leading-relaxed text-gray-700 mb-4">Vision-based Fallen Person (VFP290K) 데이터셋은 49개의 배경, 131개의 장면을 갖는 178개의 비디오로부터 낙상사고를 당한 사람에 대해 294,714개의 프레임을 추출하여 만들어졌습니다. 저희는 object detection 모델에 따른 성능 변화를 광범위한 실험을 통해 비교하였고, feature의 효과를 입증할 수 있었습니다. 또한, 저희는 낙상 감지 시스템의 성능을 측정하여 데이터세트를 평가하였습니다. 저희는 VFP290K 데이터셋을 사용하여, 2020 AI Grand Challenge의 비정상 행동 탐지 track의 첫 번째 라운드에서 1위를 달성하였고, 이는 지능형 CCTV나 감시 시스템과 같은 곳에 확대 적용될 가능성을 보여줍니다.</p>
<div class="mt-4 p-4 bg-yellow-100 rounded-lg border-l-4 border-yellow-600">
<strong class="text-gray-800">🏆 We ranked first in the first round of the anomalous behavior recognition track of AI Grand Challenge 2020, South Korea, using our VFP290K dataset, which can further extend to other applications, such as intelligent CCTV or monitoring systems, as well.</strong>
</div>
<p class="text-center mt-5">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/VFP.JPG')" alt="VFP290K Example" class="max-w-full h-auto rounded-lg shadow-lg mx-auto" />
</p>
</div>
</div>
<!-- SKKU AGC Anomaly Detection Dataset -->
<div class="section">
<div class="dash-card-dataset">
<h3 class="dash-header-dataset">📹 SKKU AGC Anomaly Detection Dataset</h3>
<p class="leading-relaxed text-gray-700 mb-6">SKKU AGC Anomaly Detection Dataset은 다양한 장소에서 낮과 밤 모두에 대해 보행자가 내려다 보이는 높이에 카메라를 고정하여 촬영하였으며 Detection Data와 Classificaiton Data로 구성되어 있습니다. Adnomal event는 사람의 머리가 땅에 닿는 경우입니다.</p>
<div class="grid md:grid-cols-2 gap-8 mb-10">
<!-- Column 1: Detection Info -->
<div class="p-5 bg-blue-50/50 rounded-xl border border-blue-100">
<h5 class="text-blue-700 font-bold mb-3 flex items-center gap-2">
<i class="fas fa-search"></i> 1. Detection Data
</h5>
<p class="text-sm leading-relaxed text-gray-700">1920x1080 크기의 이미지와 anomaly label(.xml)로 구성되어 있으며, 이미지들은 day와 night 폴더에 있으며, label들은 day_anno, night_anno 폴더에 있습니다.</p>
<div class="mt-4 flex gap-4 text-xs font-bold uppercase tracking-wider text-blue-900">
<span class="bg-white px-2 py-1 rounded shadow-sm border border-blue-200">Day: 3000</span>
<span class="bg-white px-2 py-1 rounded shadow-sm border border-blue-200">Night: 2000</span>
</div>
</div>
<!-- Column 2: Classification Info -->
<div class="p-5 bg-blue-50/50 rounded-xl border border-blue-100">
<h5 class="text-blue-700 font-bold mb-3 flex items-center gap-2">
<i class="fas fa-th-large"></i> 2. Classification Data
</h5>
<p class="text-sm leading-relaxed text-gray-700">사람을 크롭한 이미지로 구성되어 있으며, nomal과 falldown의 두가지 클래스가 있습니다. 정상 이미지는 normal day와 normal night 폴더에 있고, falldown 이미지는 falldown_day와 falldown_night 폴더에 있습니다.</p>
<div class="mt-4 grid grid-cols-2 gap-2 text-[10px] font-bold uppercase text-blue-900">
<span class="bg-white px-2 py-1 rounded shadow-sm border border-blue-200">Normal Day: 3200</span>
<span class="bg-white px-2 py-1 rounded shadow-sm border border-blue-200">Normal Night: 1300</span>
<span class="bg-white px-2 py-1 rounded shadow-sm border border-blue-200">Fall Day: 3700</span>
<span class="bg-white px-2 py-1 rounded shadow-sm border border-blue-200">Fall Night: 900</span>
</div>
</div>
</div>
<hr class="my-8 border-gray-200" />
<!-- Detection Gallery -->
<h5 class="text-gray-800 font-bold mb-4 flex items-center gap-2">
<span class="w-1.5 h-6 bg-blue-600 rounded-full"></span> Detection Examples
</h5>
<div class="grid grid-cols-2 md:grid-cols-3 gap-4 mb-12">
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-video bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_detection.jpg')" class="w-full h-full object-cover hover:scale-105 transition-transform duration-500" alt="AGC Detection Example 1" />
</div>
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-video bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_detection3.jpg')" class="w-full h-full object-cover hover:scale-105 transition-transform duration-500" alt="AGC Detection Example 2" />
</div>
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-video bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_detection4.jpg')" class="w-full h-full object-cover hover:scale-105 transition-transform duration-500" alt="AGC Detection Example 3" />
</div>
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-video bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_detection1.jpg')" class="w-full h-full object-cover hover:scale-105 transition-transform duration-500" alt="AGC Detection Example 4" />
</div>
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-video bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_detection2.jpg')" class="w-full h-full object-cover hover:scale-105 transition-transform duration-500" alt="AGC Detection Example 5" />
</div>
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-video bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_detection5.jpg')" class="w-full h-full object-cover hover:scale-105 transition-transform duration-500" alt="AGC Detection Example 6" />
</div>
</div>
<!-- Classification Gallery -->
<h5 class="text-gray-800 font-bold mb-4 flex items-center gap-2">
<span class="w-1.5 h-6 bg-blue-600 rounded-full"></span> Classification Examples
</h5>
<div class="grid grid-cols-3 sm:grid-cols-4 md:grid-cols-6 gap-3">
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-square bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_classification.jpg')" class="w-full h-full object-cover hover:scale-110 transition-transform duration-500" alt="AGC Classification Example 1" />
</div>
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-square bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_classification1.jpg')" class="w-full h-full object-cover hover:scale-110 transition-transform duration-500" alt="AGC Classification Example 2" />
</div>
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-square bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_classification2.jpg')" class="w-full h-full object-cover hover:scale-110 transition-transform duration-500" alt="AGC Classification Example 3" />
</div>
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-square bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_classification3.jpg')" class="w-full h-full object-cover hover:scale-110 transition-transform duration-500" alt="AGC Classification Example 4" />
</div>
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-square bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_classification4.jpg')" class="w-full h-full object-cover hover:scale-110 transition-transform duration-500" alt="AGC Classification Example 5" />
</div>
<div class="overflow-hidden rounded-lg shadow-sm border border-gray-200 aspect-square bg-gray-100">
<img loading="lazy" src="" onerror="this.src=getImg('img/datasets/VFP290k/AGC_classification5.jpg')" class="w-full h-full object-cover hover:scale-110 transition-transform duration-500" alt="AGC Classification Example 6" />
</div>
</div>
</div>
</div>
<!-- Deepfake Inspector PORTAL SECTION -->
<div class="section mt-12 mb-16">
<div class="dash-card-dataset overflow-hidden border-2 border-blue-200">
<div class="grid grid-cols-1 lg:grid-cols-12 gap-8 items-center">
<!-- Content Side -->
<div class="lg:col-span-7">
<div class="flex items-center gap-3 mb-4">
<span class="bg-blue-600 text-white text-[10px] font-bold px-2 py-1 rounded tracking-widest uppercase shadow-sm">Tool Beta</span>
<h3 class="text-2xl font-bold text-gray-800 m-0">🔍 Interactive Deepfake Inspector</h3>
</div>
<p class="text-gray-600 leading-relaxed mb-6 text-sm md:text-base">
데이터 세트를 제공하는 것 외에도 실시간 분석을 위한 인터랙티브 워크스테이션을 제공합니다. <strong>히스토그램 분석</strong> 도구를 사용하여 숨겨진 조작 아티팩트와 엣지 불일치를 확인합니다. 시작하려면 오른쪽에 있는 워크스테이션을 선택하세요.
</p>
<!-- Feature Grid -->
<div class="grid grid-cols-1 sm:grid-cols-3 gap-3">
<div class="bg-blue-50/50 p-3 rounded-lg flex items-center gap-3 border border-blue-100">
<i class="fas fa-chart-bar text-blue-600 text-sm"></i>
<span class="text-[10px] font-bold text-blue-900 uppercase">Histogram analysis</span>
</div>
<div class="bg-blue-50/50 p-3 rounded-lg flex items-center gap-3 border border-blue-100">
<i class="fas fa-search-plus text-blue-600 text-sm"></i>
<span class="text-[10px] font-bold text-blue-900 uppercase">Adjustable Zoom</span>
</div>
<div class="bg-blue-50/50 p-3 rounded-lg flex items-center gap-3 border border-blue-100">
<i class="fas fa-camera text-blue-600 text-sm"></i>
<span class="text-[10px] font-bold text-blue-900 uppercase">Evidence Snapshots</span>
</div>
</div>
</div>
<!-- Visual Side: Redesigned as a prominent "Action" card -->
<div class="lg:col-span-5 w-full">
<a href="Foren_ins_kor.html" class="block group">
<div class="relative bg-gray-900 rounded-2xl p-8 aspect-video flex flex-col items-center justify-center border-4 border-gray-800 shadow-2xl overflow-hidden transition-all duration-300 group-hover:border-blue-600 group-hover:shadow-blue-500/30">
<!-- Modern Scanning Line -->
<div class="absolute top-0 left-0 w-full h-1 bg-blue-500/50 shadow-[0_0_15px_rgba(59,130,246,0.8)] animate-scan z-20"></div>
<!-- Background Mesh -->
<div class="absolute inset-0 opacity-10 bg-[url('https://www.transparenttextures.com/patterns/carbon-fibre.png')]"></div>
<!-- Central Interactive UI -->
<div class="relative z-10 text-center">
<div class="w-24 h-24 rounded-full border-2 border-blue-500/30 flex items-center justify-center mb-4 mx-auto relative group-hover:scale-110 transition-transform duration-500">
<!-- Pulsing Glow -->
<div class="absolute inset-0 rounded-full bg-blue-500/20 animate-ping"></div>
<i class="fas fa-microscope text-5xl text-blue-400 group-hover:text-white transition-colors"></i>
</div>
<div class="space-y-1">
<span class="text-white text-xs font-bold uppercase tracking-[0.2em] group-hover:text-blue-400 transition-colors">Launch Station</span>
<p class="text-[9px] font-mono text-blue-300/60 uppercase tracking-widest">Histogram UI · v1.0b</p>
</div>
</div>
<!-- Clear Visual Hint -->
<div class="absolute bottom-4 left-0 w-full text-center">
<span class="text-[10px] text-gray-500 uppercase font-bold tracking-tighter animate-pulse group-hover:text-blue-200">
<i class="fas fa-mouse-pointer mr-1"></i> Click to Enter Analysis Mode
</span>
</div>
</div>
</a>
</div>
</div>
</div>
</div>
</div>
<!-- Footer is injected here by common.js -->
<style>
@keyframes spin-slow {
from { transform: rotate(0deg); }
to { transform: rotate(360deg); }
}
@keyframes scan {
0% { top: 0%; opacity: 0; }
10% { opacity: 1; }
90% { opacity: 1; }
100% { top: 100%; opacity: 0; }
}
.animate-spin-slow {
animation: spin-slow 12s linear infinite;
}
.animate-scan {
animation: scan 3s linear infinite;
}
</style>
</div>
<!-- Footer is injected here by common.js -->
</body>
</html>