Commit
·
a21dde3
1
Parent(s):
656300e
Improve UI with Tufte-inspired design principles
Browse files- Add model documentation and links to davanstrien/iconclass-vlm
- Simplify visual design: remove shadows, minimize borders, reduce clutter
- Make raw predictions collapsible (hidden by default)
- Switch to single-column layout for focused evaluation
- Improve typography: consistent sizing, lighter secondary text
- Minimize color palette: only grays and subtle green for matches
- Add CLAUDE.md documentation file
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <[email protected]>
- CLAUDE.md +46 -0
- index.html +156 -123
CLAUDE.md
ADDED
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# CLAUDE.md
|
2 |
+
|
3 |
+
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
|
4 |
+
|
5 |
+
## Project Overview
|
6 |
+
|
7 |
+
This is a Hugging Face Space that visualizes ICONCLASS predictions from the `davanstrien/iconclass-vlm` model compared to ground truth labels. It's a static web application that fetches and displays data from the `davanstrien/iconclass-sft-predictions` dataset.
|
8 |
+
|
9 |
+
The model being evaluated is a fine-tuned vision-language model (based on Qwen/Qwen2.5-VL-3B-Instruct) that automatically classifies art and cultural heritage images using Iconclass notation. The visualization shows how well the model's predictions match the ground truth labels.
|
10 |
+
|
11 |
+
## Architecture
|
12 |
+
|
13 |
+
The project consists of a single-page web application:
|
14 |
+
- `index.html`: Main application with embedded CSS and JavaScript
|
15 |
+
- `style.css`: Additional styles (currently minimal, most styles are inline in index.html)
|
16 |
+
- Uses the Hugging Face Datasets Server API to fetch data
|
17 |
+
|
18 |
+
## Key Implementation Details
|
19 |
+
|
20 |
+
### Data Source
|
21 |
+
- Dataset: `davanstrien/iconclass-sft-predictions`
|
22 |
+
- Config: `default`
|
23 |
+
- Split: `test`
|
24 |
+
- API endpoint: `https://datasets-server.huggingface.co/rows`
|
25 |
+
|
26 |
+
### Data Structure
|
27 |
+
Each row contains:
|
28 |
+
- `images`: Array of image objects with `src` URLs
|
29 |
+
- `iconclass-prediction`: Raw prediction text
|
30 |
+
- `iconclass-predictions-parsed`: Parsed prediction labels array
|
31 |
+
- `iconclass-gt-parsed`: Parsed ground truth labels array
|
32 |
+
|
33 |
+
### Core Functionality
|
34 |
+
- Lazy loading with pagination (10 images at a time)
|
35 |
+
- Infinite scroll support
|
36 |
+
- Visual comparison between predictions and ground truth
|
37 |
+
- Match detection and scoring
|
38 |
+
- Invalid label detection (labels containing "not a valid" or "invalid")
|
39 |
+
|
40 |
+
## Development Notes
|
41 |
+
|
42 |
+
Since this is a static Hugging Face Space (sdk: static), there's no build process or backend. All changes are made directly to the HTML/CSS files and are immediately reflected when deployed to Hugging Face Spaces.
|
43 |
+
|
44 |
+
**Important**: The README.md file is not displayed in static Spaces. Any documentation or description about the Space must be added directly to the `index.html` file to be visible to users.
|
45 |
+
|
46 |
+
To test locally, simply open `index.html` in a web browser.
|
index.html
CHANGED
@@ -2,7 +2,8 @@
|
|
2 |
<html>
|
3 |
<head>
|
4 |
<meta charset="UTF-8" />
|
5 |
-
<
|
|
|
6 |
<style>
|
7 |
* {
|
8 |
margin: 0;
|
@@ -13,77 +14,120 @@
|
|
13 |
body {
|
14 |
font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto,
|
15 |
sans-serif;
|
16 |
-
background:
|
17 |
padding: 20px;
|
18 |
-
line-height: 1.
|
|
|
|
|
19 |
}
|
20 |
|
21 |
.header {
|
22 |
-
|
23 |
-
margin
|
24 |
-
padding:
|
|
|
|
|
25 |
}
|
26 |
|
27 |
h1 {
|
28 |
-
font-size:
|
29 |
font-weight: 600;
|
30 |
color: #333;
|
31 |
-
margin
|
32 |
}
|
33 |
|
34 |
-
.
|
35 |
font-size: 14px;
|
36 |
color: #666;
|
|
|
|
|
37 |
}
|
38 |
|
39 |
-
.
|
40 |
-
|
41 |
-
|
42 |
-
gap: 30px;
|
43 |
-
max-width: 1200px;
|
44 |
-
margin: 0 auto;
|
45 |
}
|
46 |
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
51 |
}
|
52 |
|
53 |
.card {
|
54 |
background: white;
|
55 |
-
border
|
|
|
56 |
overflow: hidden;
|
57 |
-
|
58 |
}
|
59 |
|
60 |
.card img {
|
61 |
width: 100%;
|
62 |
height: auto;
|
|
|
|
|
63 |
display: block;
|
64 |
-
border-bottom: 1px solid #
|
|
|
65 |
}
|
66 |
|
67 |
.card-content {
|
68 |
-
padding:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
}
|
70 |
|
71 |
.raw-prediction {
|
72 |
-
|
73 |
-
|
74 |
-
|
|
|
|
|
75 |
font-family: "SF Mono", Monaco, "Cascadia Code", monospace;
|
76 |
-
font-size:
|
77 |
color: #666;
|
78 |
-
margin-bottom:
|
79 |
word-break: break-all;
|
|
|
|
|
|
|
|
|
|
|
80 |
}
|
81 |
|
82 |
.comparison {
|
83 |
display: grid;
|
84 |
grid-template-columns: 1fr 1fr;
|
85 |
gap: 20px;
|
86 |
-
margin-top:
|
87 |
}
|
88 |
|
89 |
.column {
|
@@ -91,147 +135,120 @@
|
|
91 |
}
|
92 |
|
93 |
.column-title {
|
94 |
-
font-weight:
|
95 |
margin-bottom: 8px;
|
96 |
-
color: #
|
97 |
-
font-size:
|
98 |
-
text-transform: uppercase;
|
99 |
letter-spacing: 0.5px;
|
|
|
100 |
}
|
101 |
|
102 |
.label {
|
103 |
-
padding: 6px
|
104 |
-
margin:
|
105 |
-
border-radius: 6px;
|
106 |
position: relative;
|
107 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
108 |
}
|
109 |
|
110 |
/* Prediction styles */
|
111 |
.prediction {
|
112 |
-
|
113 |
-
border: 1px solid #d0deff;
|
114 |
}
|
115 |
|
116 |
.prediction.invalid {
|
117 |
-
|
118 |
-
border: 1px solid #ffcccc;
|
119 |
-
color: #cc0000;
|
120 |
text-decoration: line-through;
|
121 |
-
opacity: 0.7;
|
122 |
}
|
123 |
|
124 |
.prediction.match {
|
125 |
-
|
126 |
-
|
127 |
-
}
|
128 |
-
|
129 |
-
.prediction.match::before {
|
130 |
-
content: "✓";
|
131 |
-
position: absolute;
|
132 |
-
left: -20px;
|
133 |
-
color: #4caf50;
|
134 |
-
font-weight: bold;
|
135 |
}
|
136 |
|
137 |
/* Ground truth styles */
|
138 |
.ground-truth {
|
139 |
-
|
140 |
-
border: 1px solid #e0e0e0;
|
141 |
}
|
142 |
|
143 |
.ground-truth.matched {
|
144 |
-
|
145 |
-
|
146 |
-
}
|
147 |
-
|
148 |
-
/* Match
|
149 |
-
.match-
|
150 |
-
|
151 |
-
|
152 |
-
|
153 |
-
|
154 |
-
|
155 |
-
border-radius: 10px;
|
156 |
-
font-size: 10px;
|
157 |
-
font-weight: 600;
|
158 |
}
|
159 |
|
160 |
.controls {
|
161 |
text-align: center;
|
162 |
-
margin:
|
|
|
163 |
}
|
164 |
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
padding: 12px 30px;
|
170 |
-
border-radius: 8px;
|
171 |
-
font-size: 14px;
|
172 |
cursor: pointer;
|
173 |
-
|
174 |
-
|
|
|
|
|
175 |
}
|
176 |
|
177 |
-
|
178 |
-
|
179 |
-
transform: translateY(-1px);
|
180 |
}
|
181 |
|
182 |
-
|
183 |
-
|
184 |
-
cursor:
|
185 |
-
|
186 |
}
|
187 |
|
188 |
.loading {
|
189 |
text-align: center;
|
190 |
-
padding:
|
191 |
-
color: #
|
192 |
-
font-size:
|
193 |
}
|
194 |
|
195 |
.loading.hidden {
|
196 |
display: none;
|
197 |
}
|
198 |
-
|
199 |
-
.score-badge {
|
200 |
-
display: inline-block;
|
201 |
-
margin-top: 10px;
|
202 |
-
padding: 4px 12px;
|
203 |
-
background: #f0f0f0;
|
204 |
-
border-radius: 20px;
|
205 |
-
font-size: 12px;
|
206 |
-
color: #666;
|
207 |
-
}
|
208 |
-
|
209 |
-
.score-badge.good {
|
210 |
-
background: #e8f5e9;
|
211 |
-
color: #2e7d32;
|
212 |
-
}
|
213 |
-
|
214 |
-
.score-badge.poor {
|
215 |
-
background: #fff3e0;
|
216 |
-
color: #e65100;
|
217 |
-
}
|
218 |
</style>
|
219 |
</head>
|
220 |
<body>
|
221 |
<div class="header">
|
222 |
-
<h1>ICONCLASS
|
|
|
|
|
|
|
|
|
|
|
|
|
223 |
<div class="stats">
|
224 |
-
<span id="loadedCount">0</span>
|
225 |
-
loaded
|
226 |
</div>
|
227 |
</div>
|
228 |
|
229 |
<div id="gallery" class="gallery"></div>
|
230 |
|
231 |
-
<div class="loading hidden" id="loading">Loading
|
232 |
|
233 |
<div class="controls">
|
234 |
-
<button id="loadMore">Load
|
235 |
</div>
|
236 |
|
237 |
<script>
|
@@ -283,7 +300,7 @@
|
|
283 |
loadMoreBtn.disabled = true;
|
284 |
loadMoreBtn.textContent = "All Images Loaded";
|
285 |
} else {
|
286 |
-
loadMoreBtn.textContent = `Load
|
287 |
totalRows - currentOffset
|
288 |
} remaining)`;
|
289 |
}
|
@@ -317,11 +334,27 @@
|
|
317 |
const content = document.createElement("div");
|
318 |
content.className = "card-content";
|
319 |
|
320 |
-
// Show raw prediction
|
321 |
if (row["iconclass-prediction"]) {
|
|
|
|
|
|
|
|
|
322 |
const rawDiv = document.createElement("div");
|
323 |
rawDiv.className = "raw-prediction";
|
324 |
rawDiv.textContent = row["iconclass-prediction"];
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
325 |
content.appendChild(rawDiv);
|
326 |
}
|
327 |
|
@@ -399,7 +432,7 @@
|
|
399 |
comparison.appendChild(gtColumn);
|
400 |
content.appendChild(comparison);
|
401 |
|
402 |
-
// Add match
|
403 |
const validPredictions = predictions.filter(
|
404 |
(_, idx) => !invalidPredictions[idx]
|
405 |
);
|
@@ -408,12 +441,12 @@
|
|
408 |
? Math.round((matches.length / validPredictions.length) * 100)
|
409 |
: 0;
|
410 |
|
411 |
-
const
|
412 |
-
|
413 |
-
|
414 |
-
|
415 |
-
|
416 |
-
content.appendChild(
|
417 |
|
418 |
card.appendChild(content);
|
419 |
gallery.appendChild(card);
|
|
|
2 |
<html>
|
3 |
<head>
|
4 |
<meta charset="UTF-8" />
|
5 |
+
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
|
6 |
+
<title>ICONCLASS Model Evaluation - davanstrien/iconclass-vlm</title>
|
7 |
<style>
|
8 |
* {
|
9 |
margin: 0;
|
|
|
14 |
body {
|
15 |
font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto,
|
16 |
sans-serif;
|
17 |
+
background: white;
|
18 |
padding: 20px;
|
19 |
+
line-height: 1.5;
|
20 |
+
color: #333;
|
21 |
+
font-size: 14px;
|
22 |
}
|
23 |
|
24 |
.header {
|
25 |
+
max-width: 800px;
|
26 |
+
margin: 0 auto 30px;
|
27 |
+
padding: 0;
|
28 |
+
border-bottom: 1px solid #e5e5e5;
|
29 |
+
padding-bottom: 20px;
|
30 |
}
|
31 |
|
32 |
h1 {
|
33 |
+
font-size: 20px;
|
34 |
font-weight: 600;
|
35 |
color: #333;
|
36 |
+
margin: 0 0 8px 0;
|
37 |
}
|
38 |
|
39 |
+
.subtitle {
|
40 |
font-size: 14px;
|
41 |
color: #666;
|
42 |
+
margin: 0 0 12px 0;
|
43 |
+
line-height: 1.4;
|
44 |
}
|
45 |
|
46 |
+
.subtitle a {
|
47 |
+
color: #0066cc;
|
48 |
+
text-decoration: none;
|
|
|
|
|
|
|
49 |
}
|
50 |
|
51 |
+
.subtitle a:hover {
|
52 |
+
text-decoration: underline;
|
53 |
+
}
|
54 |
+
|
55 |
+
.description {
|
56 |
+
font-size: 13px;
|
57 |
+
color: #666;
|
58 |
+
margin: 0 0 8px 0;
|
59 |
+
line-height: 1.5;
|
60 |
+
}
|
61 |
+
|
62 |
+
.stats {
|
63 |
+
font-size: 11px;
|
64 |
+
color: #999;
|
65 |
+
margin: 0;
|
66 |
+
}
|
67 |
+
|
68 |
+
.gallery {
|
69 |
+
max-width: 800px;
|
70 |
+
margin: 0 auto;
|
71 |
}
|
72 |
|
73 |
.card {
|
74 |
background: white;
|
75 |
+
border: 1px solid #e5e5e5;
|
76 |
+
border-radius: 2px;
|
77 |
overflow: hidden;
|
78 |
+
margin-bottom: 20px;
|
79 |
}
|
80 |
|
81 |
.card img {
|
82 |
width: 100%;
|
83 |
height: auto;
|
84 |
+
max-height: 500px;
|
85 |
+
object-fit: contain;
|
86 |
display: block;
|
87 |
+
border-bottom: 1px solid #e5e5e5;
|
88 |
+
background: #fafafa;
|
89 |
}
|
90 |
|
91 |
.card-content {
|
92 |
+
padding: 15px;
|
93 |
+
}
|
94 |
+
|
95 |
+
.raw-toggle {
|
96 |
+
font-size: 11px;
|
97 |
+
color: #999;
|
98 |
+
cursor: pointer;
|
99 |
+
margin-bottom: 12px;
|
100 |
+
user-select: none;
|
101 |
+
font-family: "SF Mono", Monaco, "Cascadia Code", monospace;
|
102 |
+
}
|
103 |
+
|
104 |
+
.raw-toggle:hover {
|
105 |
+
color: #666;
|
106 |
}
|
107 |
|
108 |
.raw-prediction {
|
109 |
+
display: none;
|
110 |
+
background: #fafafa;
|
111 |
+
padding: 8px;
|
112 |
+
border: 1px solid #e5e5e5;
|
113 |
+
border-radius: 2px;
|
114 |
font-family: "SF Mono", Monaco, "Cascadia Code", monospace;
|
115 |
+
font-size: 11px;
|
116 |
color: #666;
|
117 |
+
margin-bottom: 15px;
|
118 |
word-break: break-all;
|
119 |
+
line-height: 1.4;
|
120 |
+
}
|
121 |
+
|
122 |
+
.raw-prediction.visible {
|
123 |
+
display: block;
|
124 |
}
|
125 |
|
126 |
.comparison {
|
127 |
display: grid;
|
128 |
grid-template-columns: 1fr 1fr;
|
129 |
gap: 20px;
|
130 |
+
margin-top: 0;
|
131 |
}
|
132 |
|
133 |
.column {
|
|
|
135 |
}
|
136 |
|
137 |
.column-title {
|
138 |
+
font-weight: 400;
|
139 |
margin-bottom: 8px;
|
140 |
+
color: #999;
|
141 |
+
font-size: 10px;
|
|
|
142 |
letter-spacing: 0.5px;
|
143 |
+
text-transform: uppercase;
|
144 |
}
|
145 |
|
146 |
.label {
|
147 |
+
padding: 3px 0 3px 6px;
|
148 |
+
margin: 2px 0;
|
|
|
149 |
position: relative;
|
150 |
+
border-left: 2px solid transparent;
|
151 |
+
font-size: 13px;
|
152 |
+
line-height: 1.4;
|
153 |
+
}
|
154 |
+
|
155 |
+
.label:hover {
|
156 |
+
background: #fafafa;
|
157 |
}
|
158 |
|
159 |
/* Prediction styles */
|
160 |
.prediction {
|
161 |
+
color: #333;
|
|
|
162 |
}
|
163 |
|
164 |
.prediction.invalid {
|
165 |
+
color: #999;
|
|
|
|
|
166 |
text-decoration: line-through;
|
|
|
167 |
}
|
168 |
|
169 |
.prediction.match {
|
170 |
+
border-left-color: #4caf50;
|
171 |
+
font-weight: 500;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
172 |
}
|
173 |
|
174 |
/* Ground truth styles */
|
175 |
.ground-truth {
|
176 |
+
color: #333;
|
|
|
177 |
}
|
178 |
|
179 |
.ground-truth.matched {
|
180 |
+
border-left-color: #4caf50;
|
181 |
+
font-weight: 500;
|
182 |
+
}
|
183 |
+
|
184 |
+
/* Match statistics */
|
185 |
+
.match-stats {
|
186 |
+
font-size: 11px;
|
187 |
+
color: #999;
|
188 |
+
margin-top: 10px;
|
189 |
+
padding-top: 10px;
|
190 |
+
border-top: 1px solid #f0f0f0;
|
|
|
|
|
|
|
191 |
}
|
192 |
|
193 |
.controls {
|
194 |
text-align: center;
|
195 |
+
margin: 30px 0;
|
196 |
+
padding-top: 20px;
|
197 |
}
|
198 |
|
199 |
+
.load-more {
|
200 |
+
color: #0066cc;
|
201 |
+
text-decoration: none;
|
202 |
+
font-size: 13px;
|
|
|
|
|
|
|
203 |
cursor: pointer;
|
204 |
+
background: none;
|
205 |
+
border: none;
|
206 |
+
padding: 0;
|
207 |
+
font-family: inherit;
|
208 |
}
|
209 |
|
210 |
+
.load-more:hover {
|
211 |
+
text-decoration: underline;
|
|
|
212 |
}
|
213 |
|
214 |
+
.load-more:disabled {
|
215 |
+
color: #999;
|
216 |
+
cursor: default;
|
217 |
+
text-decoration: none;
|
218 |
}
|
219 |
|
220 |
.loading {
|
221 |
text-align: center;
|
222 |
+
padding: 10px;
|
223 |
+
color: #999;
|
224 |
+
font-size: 12px;
|
225 |
}
|
226 |
|
227 |
.loading.hidden {
|
228 |
display: none;
|
229 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
230 |
</style>
|
231 |
</head>
|
232 |
<body>
|
233 |
<div class="header">
|
234 |
+
<h1>ICONCLASS Model Evaluation</h1>
|
235 |
+
<div class="subtitle">
|
236 |
+
Comparing predictions from <a href="https://huggingface.co/davanstrien/iconclass-vlm" target="_blank">davanstrien/iconclass-vlm</a> against ground truth labels
|
237 |
+
</div>
|
238 |
+
<div class="description">
|
239 |
+
A vision-language model fine-tuned on Qwen2.5-VL-3B for classifying art and cultural heritage images using ICONCLASS notation — a hierarchical classification system for art and iconography.
|
240 |
+
</div>
|
241 |
<div class="stats">
|
242 |
+
Showing <span id="loadedCount">0</span> of <span id="totalCount">-</span> test images
|
|
|
243 |
</div>
|
244 |
</div>
|
245 |
|
246 |
<div id="gallery" class="gallery"></div>
|
247 |
|
248 |
+
<div class="loading hidden" id="loading">Loading...</div>
|
249 |
|
250 |
<div class="controls">
|
251 |
+
<button id="loadMore" class="load-more">Load more images</button>
|
252 |
</div>
|
253 |
|
254 |
<script>
|
|
|
300 |
loadMoreBtn.disabled = true;
|
301 |
loadMoreBtn.textContent = "All Images Loaded";
|
302 |
} else {
|
303 |
+
loadMoreBtn.textContent = `Load more images (${
|
304 |
totalRows - currentOffset
|
305 |
} remaining)`;
|
306 |
}
|
|
|
334 |
const content = document.createElement("div");
|
335 |
content.className = "card-content";
|
336 |
|
337 |
+
// Show raw prediction (collapsible)
|
338 |
if (row["iconclass-prediction"]) {
|
339 |
+
const toggleDiv = document.createElement("div");
|
340 |
+
toggleDiv.className = "raw-toggle";
|
341 |
+
toggleDiv.textContent = "+ Show raw prediction";
|
342 |
+
|
343 |
const rawDiv = document.createElement("div");
|
344 |
rawDiv.className = "raw-prediction";
|
345 |
rawDiv.textContent = row["iconclass-prediction"];
|
346 |
+
|
347 |
+
toggleDiv.addEventListener("click", () => {
|
348 |
+
if (rawDiv.classList.contains("visible")) {
|
349 |
+
rawDiv.classList.remove("visible");
|
350 |
+
toggleDiv.textContent = "+ Show raw prediction";
|
351 |
+
} else {
|
352 |
+
rawDiv.classList.add("visible");
|
353 |
+
toggleDiv.textContent = "− Hide raw prediction";
|
354 |
+
}
|
355 |
+
});
|
356 |
+
|
357 |
+
content.appendChild(toggleDiv);
|
358 |
content.appendChild(rawDiv);
|
359 |
}
|
360 |
|
|
|
432 |
comparison.appendChild(gtColumn);
|
433 |
content.appendChild(comparison);
|
434 |
|
435 |
+
// Add match statistics
|
436 |
const validPredictions = predictions.filter(
|
437 |
(_, idx) => !invalidPredictions[idx]
|
438 |
);
|
|
|
441 |
? Math.round((matches.length / validPredictions.length) * 100)
|
442 |
: 0;
|
443 |
|
444 |
+
const statsDiv = document.createElement("div");
|
445 |
+
statsDiv.className = "match-stats";
|
446 |
+
statsDiv.textContent = validPredictions.length > 0
|
447 |
+
? `${matches.length}/${validPredictions.length} matches`
|
448 |
+
: `No valid predictions`;
|
449 |
+
content.appendChild(statsDiv);
|
450 |
|
451 |
card.appendChild(content);
|
452 |
gallery.appendChild(card);
|