-
Notifications
You must be signed in to change notification settings - Fork 7
Expand file tree
/
Copy pathtask-gsr.html
More file actions
426 lines (381 loc) · 13.8 KB
/
task-gsr.html
File metadata and controls
426 lines (381 loc) · 13.8 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
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>GSR Task - SoccerTrack v2</title>
<style>
* {
margin: 0;
padding: 0;
box-sizing: border-box;
}
body {
font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Oxygen, Ubuntu, Cantarell, sans-serif;
line-height: 1.6;
color: #333;
background: linear-gradient(135deg, #1a1a1a 0%, #0a0a0a 100%);
min-height: 100vh;
padding: 2rem;
}
.container {
max-width: 900px;
margin: 0 auto;
background: #ffffff;
border-radius: 12px;
box-shadow: 0 20px 60px rgba(16, 185, 129, 0.2);
overflow: hidden;
}
header {
background: #1a1a1a;
color: white;
padding: 2rem;
border-bottom: 3px solid #10b981;
}
h1 {
font-size: 2rem;
margin-bottom: 0.5rem;
}
.breadcrumb {
font-size: 0.9rem;
opacity: 0.8;
}
.breadcrumb a {
color: #10b981;
text-decoration: none;
}
.breadcrumb a:hover {
text-decoration: underline;
}
main {
padding: 2rem;
}
h2 {
color: #10b981;
margin-top: 2rem;
margin-bottom: 1rem;
font-size: 1.5rem;
border-bottom: 2px solid #10b981;
padding-bottom: 0.5rem;
}
h2:first-child {
margin-top: 0;
}
h3 {
color: #059669;
margin-top: 1.5rem;
margin-bottom: 0.5rem;
}
table {
width: 100%;
border-collapse: collapse;
margin: 1rem 0;
background: #f8f9fa;
border-radius: 8px;
overflow: hidden;
}
th, td {
padding: 0.75rem;
text-align: left;
border-bottom: 1px solid #e0e0e0;
}
th {
background: #10b981;
color: white;
font-weight: 600;
}
tr:last-child td {
border-bottom: none;
}
code {
background: #f4f4f4;
padding: 0.2rem 0.5rem;
border-radius: 4px;
font-family: 'Courier New', monospace;
font-size: 0.9rem;
}
pre {
background: #2d2d2d;
color: #f8f8f2;
padding: 1.5rem;
border-radius: 8px;
overflow-x: auto;
margin: 1rem 0;
}
pre code {
background: none;
color: inherit;
padding: 0;
}
.highlight-box {
background: #f0fdf4;
border-left: 4px solid #10b981;
padding: 1rem;
margin: 1rem 0;
border-radius: 4px;
}
.diff-box {
background: #fffbeb;
border-left: 4px solid #f59e0b;
padding: 1rem;
margin: 1rem 0;
border-radius: 4px;
}
.kit-box {
background: #ecfdf5;
border-left: 4px solid #10b981;
padding: 1rem 1.25rem;
margin: 1rem 0;
border-radius: 4px;
}
.kit-box .btn-row {
margin-top: 0.75rem;
display: flex;
gap: 0.5rem;
flex-wrap: wrap;
}
.kit-box a.btn-kit {
display: inline-block;
padding: 0.5rem 1rem;
background: #10b981;
color: white;
border-radius: 6px;
text-decoration: none;
font-weight: 600;
}
.kit-box a.btn-kit.alt {
background: #fff;
color: #10b981;
border: 1px solid #10b981;
}
ul {
margin-left: 2rem;
margin-top: 0.5rem;
}
li {
margin: 0.5rem 0;
}
footer {
background: #2d2d2d;
color: white;
padding: 1.5rem 2rem;
text-align: center;
font-size: 0.9rem;
}
footer a {
color: #10b981;
text-decoration: none;
}
footer a:hover {
text-decoration: underline;
}
</style>
</head>
<body>
<div class="container">
<header>
<p class="breadcrumb"><a href="index.html">SoccerTrack v2</a> / Tasks</p>
<h1>🗺️ Game State Reconstruction (GSR)</h1>
</header>
<main>
<section id="overview">
<h2>Task Overview</h2>
<p>
<strong>Input:</strong> Panoramic 4K video (full match)<br>
<strong>Output:</strong> Per-frame game state on 2D pitch minimap (position + role + team + jersey number)
</p>
<div class="highlight-box">
<strong>Alignment with SoccerNet-GSR:</strong><br>
✅ <strong>Same:</strong> 2D pitch coordinates, player roles, team assignment, jersey numbers<br>
⚠️ <strong>Different:</strong> Fixed panoramic camera + full match (not broadcast clips)
</div>
</section>
<section id="io">
<h2>Input / Output Definition</h2>
<h3>Input</h3>
<ul>
<li><code>videos/match_XXXXX.mp4</code> — Full match panoramic video (4K)</li>
<li>Camera: BePro Cerberus (2 matches) or 3-camera panoramic (8 matches)</li>
</ul>
<h3>Output (per frame)</h3>
<table>
<thead>
<tr>
<th>Field</th>
<th>Type</th>
<th>Description</th>
</tr>
</thead>
<tbody>
<tr>
<td><code>frame</code></td>
<td>int</td>
<td>Frame number</td>
</tr>
<tr>
<td><code>time</code></td>
<td>float</td>
<td>Timestamp in video (seconds)</td>
</tr>
<tr>
<td><code>player_id</code></td>
<td>int</td>
<td>Unique track ID (persistent throughout match)</td>
</tr>
<tr>
<td><code>x, y</code></td>
<td>float</td>
<td>2D pitch coordinates in meters</td>
</tr>
<tr>
<td><code>role</code></td>
<td>string</td>
<td>player / goalkeeper / referee / other</td>
</tr>
<tr>
<td><code>team_side</code></td>
<td>string</td>
<td>left / right / null</td>
</tr>
<tr>
<td><code>jersey_number</code></td>
<td>int/null</td>
<td>0–99 or null</td>
</tr>
</tbody>
</table>
</section>
<section id="layout">
<h2>File Layout</h2>
<pre><code>SoccerTrack-v2/
├── videos/
│ ├── 117093/
│ │ ├── 117093_panorama_1st_half.mp4
│ │ └── 117093_panorama_2nd_half.mp4
│ └── ...
├── gsr/
│ ├── 117093/
│ │ ├── 117093_1st.json
│ │ └── 117093_2nd.json
│ └── ...
├── bas/
├── mot/
└── raw/
├── 117093/
│ ├── 117093_keypoints.json
│ ├── 117093_mapx.npy
│ ├── 117093_mapy.npy
│ └── ...
└── ...</code></pre>
<p>
Each match ships <strong>one GSR annotation file per half</strong>. Unlike SoccerNet-GSR which stores image frames under <code>img1/</code>, we reference the trimmed half videos directly from <code>videos/</code> to save storage space.
</p>
<p>
See <a href="format-gsr.md">format-gsr.md</a> for the full schema reference and pitch coordinate system.
</p>
</section>
<section id="schema">
<h2>Schema & SoccerNet Mapping</h2>
<h3>Example JSON Record</h3>
<p>Each half's file is a flat JSON array of per-entity-per-frame records. One record:</p>
<pre><code>{
"image_id": 12480,
"track_id": 7,
"player_id": "117092_L_9",
"role": "player",
"jersey_number": 9,
"team_side": "left",
"x": 48.21,
"y": 34.07,
"bbox_image": [1840, 710, 108, 242],
"bbox_pitch": [46.8, 33.1, 1.4, 2.0]
}</code></pre>
<h3>SoccerNet-GSR Correspondence</h3>
<table>
<thead>
<tr>
<th>SoccerTrack v2</th>
<th>SoccerNet-GSR</th>
<th>Notes</th>
</tr>
</thead>
<tbody>
<tr>
<td><code>x, y</code></td>
<td><code>x_pos, y_pos</code></td>
<td>Pitch coordinates (meters)</td>
</tr>
<tr>
<td><code>team_side</code></td>
<td><code>left / right</code></td>
<td>Team assignment</td>
</tr>
<tr>
<td><code>jersey_number</code></td>
<td><code>jersey_number</code></td>
<td>0–99</td>
</tr>
<tr>
<td><code>role</code></td>
<td><code>role</code></td>
<td>player/goalkeeper/referee/other</td>
</tr>
<tr>
<td><code>player_id</code></td>
<td><code>track_id</code></td>
<td>Persistent across full match</td>
</tr>
</tbody>
</table>
</section>
<section id="evaluation">
<h2>Evaluation / Metrics</h2>
<p>
Evaluation uses <strong>GS-HOTA</strong> (Game State Higher Order Tracking Accuracy),
following the SoccerNet-GSR Challenge definition. Report GS-HOTA plus the
DetA / AssA / LocA decomposition.
</p>
<p>
Reference CLI:
</p>
<pre><code>python -m src.evaluation.gs_hota \
--pred preds/gsr --gt data/gsr --matches 117099 117100</code></pre>
<p>
The module is a thin wrapper over the upstream
<a href="https://github.com/SoccerNet/sn-gamestate">sn-gamestate</a> scorer.
</p>
</section>
<section id="starter-kit">
<h2>Starter Kit</h2>
<div class="kit-box">
<strong>Get running fast.</strong> The GSR starter kit scaffolds
detection → tracking → jersey/role tagging → pitch projection into a runnable
pipeline on one GPU. Fill in the model calls, then run eval.
<div class="btn-row">
<a class="btn-kit" href="https://github.com/AtomScott/SoccerTrack-v2/tree/main/baselines/gsr">📦 baselines/gsr/</a>
<a class="btn-kit alt" href="leaderboard.html">🏁 Leaderboard</a>
</div>
</div>
</section>
<section id="differences">
<h2>Differences vs SoccerNet-GSR</h2>
<div class="diff-box">
<strong>Key Differences:</strong>
<ul>
<li><strong>Camera view:</strong> Panoramic (full-pitch) vs. broadcast (partial view with occlusions)</li>
<li><strong>Duration:</strong> Full match (~90 min) vs. short clips</li>
<li><strong>Camera setup:</strong> Fixed BePro/3-camera system vs. moving broadcast camera</li>
<li><strong>Level:</strong> University amateur matches vs. professional leagues</li>
</ul>
</div>
</section>
</main>
<footer>
<p>
<a href="index.html">← Back to main page</a> |
<a href="https://www.soccer-net.org/tasks/game-state-reconstruction">SoccerNet-GSR Challenge</a>
</p>
</footer>
</div>
</body>
</html>