You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: LucidNFT/index.html
+28-20Lines changed: 28 additions & 20 deletions
Original file line number
Diff line number
Diff line change
@@ -77,6 +77,7 @@ <h2>Method Overview</h2>
77
77
</div>
78
78
<divclass="table-wrapper">
79
79
<tableclass="table-consistency">
80
+
<caption>Cosine similarity between LR-HR image pairs under correct pairing and cross-bench swapped pairing. LucidConsistency improves alignment on correctly paired data while maintaining low similarity under mismatched pairing.</caption>
80
81
<thead>
81
82
<tr>
82
83
<th>Domain</th>
@@ -116,9 +117,6 @@ <h2>Method Overview</h2>
116
117
</tr>
117
118
</tbody>
118
119
</table>
119
-
<divclass="table-caption">
120
-
<spanclass="table-number">Table 2:</span> Cosine similarity between LR-HR image pairs under correct pairing and cross-bench swapped pairing. LucidConsistency improves alignment on correctly paired data while maintaining low similarity under mismatched pairing.
121
-
</div>
122
120
</div>
123
121
</section>
124
122
@@ -131,6 +129,7 @@ <h2>LucidLR Dataset</h2>
131
129
</div>
132
130
<divclass="table-wrapper">
133
131
<tableclass="table-dataset">
132
+
<caption>Comparison of representative real-world datasets used in Real-ISR. Existing datasets are mainly designed for benchmarking and contain limited samples, while LucidLR provides large-scale real-world degradations suitable for preference-optimization-based training.</caption>
134
133
<thead>
135
134
<tr>
136
135
<th>Dataset</th>
@@ -171,9 +170,6 @@ <h2>LucidLR Dataset</h2>
171
170
</tr>
172
171
</tbody>
173
172
</table>
174
-
<divclass="table-caption">
175
-
<spanclass="table-number">Table 1:</span> Comparison of representative real-world datasets used in Real-ISR. Existing datasets are mainly designed for benchmarking and contain limited samples, while LucidLR provides large-scale real-world degradations suitable for preference-optimization-based training.
176
-
</div>
177
173
</div>
178
174
</section>
179
175
@@ -200,6 +196,7 @@ <h2>Results</h2>
200
196
<h3>Quantitative Results</h3>
201
197
<divclass="table-container">
202
198
<table>
199
+
<caption>Quantitative comparison with state-of-the-art Real-ISR methods on RealLQ250, DRealSR, and RealSR. Higher is better for all metrics except NIQE. Values in the last column show improvements relative to the LucidFlux baseline.</caption>
203
200
<thead>
204
201
<tr>
205
202
<throwspan="2">Benchmark</th>
@@ -626,9 +623,6 @@ <h3>Quantitative Results</h3>
626
623
</tbody>
627
624
</table>
628
625
</div>
629
-
<divclass="table-caption">
630
-
<spanclass="table-number">Table 3:</span> Quantitative comparison with state-of-the-art Real-ISR methods on RealLQ250, DRealSR, and RealSR. Higher is better for all metrics except NIQE. Values in the last column show improvements relative to the LucidFlux baseline.
0 commit comments