GeoExplorer/index.html at main · limirs/GeoExplorer · GitHub

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
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475

<!DOCTYPE html>
<html>
<head>
  <meta charset="utf-8">
  <!-- Meta tags for social media banners, these should be filled in appropriatly as they are your "business card" -->
  <!-- Replace the content tag with appropriate information -->
  <meta name="description" content="GeoExplorer: Active Geo-localization with Curiosity-Driven Exploration">
  <meta property="og:title" content="GeoExplorer: Active Geo-localization with Curiosity-Driven Exploration"/>
  <meta property="og:description" content="GeoExplorer: Active Geo-localization with Curiosity-Driven Exploration"/>
  <meta property="og:url" content=""/>
  <!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X630-->
  <meta property="og:image" content="static/image/your_banner_image.png" />
  <meta property="og:image:width" content="1200"/>
  <meta property="og:image:height" content="630"/>


  <meta name="twitter:title" content="GeoExplorer: Active Geo-localization with Curiosity-Driven Exploration">
  <meta name="twitter:description" content="GeoExplorer: Active Geo-localization with Curiosity-Driven Exploration">
  <!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X600-->
  <meta name="twitter:image" content="static/images/your_twitter_banner_image.png">
  <meta name="twitter:card" content="summary_large_image">
  <!-- Keywords for your paper to be indexed by-->
  <meta name="keywords" content="geo-localiztion, curiosity-driven reinforcement learning">
  <meta name="viewport" content="width=device-width, initial-scale=1">


  <title>GeoExplorer</title>
  <link rel="icon" type="image/x-icon" href="static/images/explore.png">
  <link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro"
  rel="stylesheet">

  <link rel="stylesheet" href="static/css/bulma.min.css">
  <link rel="stylesheet" href="static/css/bulma-carousel.min.css">
  <link rel="stylesheet" href="static/css/bulma-slider.min.css">
  <link rel="stylesheet" href="static/css/fontawesome.all.min.css">
  <link rel="stylesheet"
  href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
  <link rel="stylesheet" href="static/css/index.css">

  <script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
  <script src="https://documentcloud.adobe.com/view-sdk/main.js"></script>
  <script defer src="static/js/fontawesome.all.min.js"></script>
  <script src="static/js/bulma-carousel.min.js"></script>
  <script src="static/js/bulma-slider.min.js"></script>
  <script src="static/js/index.js"></script>
  <script type="text/x-mathjax-config"> MathJax.Hub.Config({ tex2jax: {inlineMath: [['$','$'],['\\(','\\)']]} }); </script> <script type="text/javascript" async src="https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-MML-AM_CHTML"> </script>
</head>
<body>


  <section class="hero">
    <div class="hero-body">
      <div class="container is-max-desktop">
        <div class="columns is-centered">
          <div class="column has-text-centered">
            <div style="display: flex; align-items: center; justify-content: center;">
              <img src="static/images/explore.png" style="height: 60px; margin-right: 10px; margin-top: -10px;">
              <h1 class="title is-1 publication-title"><b>GeoExplorer:</b></h1>
            </div>
            <h2 class="title is-3 has-text-centered"><b>Active Geo-localization with Curiosity-Driven Exploration</b></h2>
            <div class="is-size-5 publication-authors">
              <!-- Paper authors -->
              <span class="author-block">
                <a href="https://limirs.github.io/" target="_blank">Li Mi</a></sup>,</span>
              <span class="author-block">
                <a href="https://people.epfl.ch/manon.bechaz?lang=en" target="_blank">Manon Béchaz</a></sup>,</span>
              <span class="author-block">
                <a href="https://eric11eca.github.io/" target="_blank">Zeming Chen</a></sup>,</span>
              <span class="author-block">
                <a href="https://atcbosselut.github.io/" target="_blank">Antoine Bosselut</a>,</span>
              <span class="author-block">
                <a href="https://people.epfl.ch/devis.tuia" target="_blank">Devis Tuia</a></span>
              </div>

                  <div class="is-size-5 publication-authors">
                    <span class="author-block">EPFL, Switzerland</span>
                  </div>

                  <div class="is-size-5 publication-authors">
                    <span class="author-block"><b>ICCV 2025</b></span>
                  </div>

                  <div class="is-size-5 publication-authors">
                    <span class="author-block">
                      <p><br></p>
                      <a href="https://www.epfl.ch/en/"><img src="static/images/Fig_EPFL_Red.png" width="150px"  margin-left="20px"  margin-right="20px" alt="EPFL Logo"/></a></span>
                      <a href="http://www.epfl.ch/labs/eceo/"><img src="static/images/Fig_ECEO.png" width="150px" margin-left="20px" margin-right="20px" alt="ECEO Logo"/></a></span>
                      <a href="https://nlp.epfl.ch/"><img src="static/images/Fig_NLP.png" width="70px"  margin-left="20px" margin-right="20px" alt="NLP Logo"/></a></span>
                  </div>

                  <!---TODO ECEO LOGO, and change logo from red/white invert, put pipeline image-->
                 <div class="column has-text-centered">
                    <div class="publication-links">
                   <!-- Arxiv PDF link -->
                      <span class="link-block">
                        <a href="https://arxiv.org/abs/2508.00152" target="_blank"
                        class="external-link button is-normal is-rounded is-dark">
                        <span class="icon">
                          <i class="fas fa-file-pdf"></i>
                        </span>
                        <span>Pre-Print</span>
                      </a>
                    </span>

                  <!-- Github link -->
                  <span class="link-block">
                    <a href="https://github.com/limirs/GeoExplorer" target="_blank"
                    class="external-link button is-normal is-rounded is-dark">
                    <span class="icon">
                      <i class="fab fa-github"></i>
                    </span>
                    <span>Code</span>
                  </a>
                  </span>

                  <!-- Dataset link -->
                  <span class="link-block">
                    <a href="https://huggingface.co/datasets/EPFL-ECEO/SwissView" target="_blank"
                    class="external-link button is-normal is-rounded is-dark">
                    <span class="icon">
                    <img src="https://huggingface.co/front/assets/huggingface_logo-noborder.svg"
                    alt="Hugging Face"
                    width="20" height="20">
                    </span>
                    <span>Dataset</span>
                    </a>
                  </span>

                <div class="container is-max-desktop">
                <div class="columns is-centered">
                <div class="column is-four-fifths">
                <div class="content has-text-justified">
                  <p><br></p>
                <p>
                <div align="center">
                <b style="font-size:21px; ">GeoExplorer combines
                  <span style="text-shadow: 2px 2px 4px rgba(0, 0, 0, 0.5);">
                    goal-oriented reward
                  </span>
                  and
                  <span style="text-shadow: 2px 2px 4px rgba(255, 0, 0, 0.5); color: #792714">
                    curiosity-driven reward
                  </span>
                  <br>
                to address the task of Active Geo-localization</b>
                <p><br></p>
                </div>
                <div align="center">
                <img src="static/images/Fig_Summary.png" alt="An Overview of ConGeo" width="80%"/>
                </div>
                Integrating curiosity-driven rewards with goal-oriented rewards introduces an essential trade-off between
                <span style="text-shadow: 2px 2px 4px rgba(0, 0, 0, 0.5);">
                    following the direct, goal-oriented guidance
                </span>
                and
                <span style="text-shadow: 2px 2px 4px rgba(255, 0, 0, 0.5); color: #792714">
                    engaging in exploratory behavior.
                </span>
            </div>
          </div>
        </div>
      </div>
    </div>
  </div>
</section>


<!-- Paper abstract -->
<section class="section hero is-light">
  <div class="container is-max-desktop">
    <div class="columns is-centered">
      <div class="column is-four-fifths">
        <h2 class="title is-3 has-text-centered">Active Geo-localization</h2>
        <div class="content has-text-justified">
        <p><b>Active Geo-Localization (AGL)</b> within a goal-reaching <b>reinforcement learning (RL)</b> context.<br>
          <b>(a)</b> AGL focuses on localizing a target (<b>goal</b>), within a predefined search area (<b>environment</b>) presented in the bird’s eye view, by navigating the agent towards it. At a given time, the agent observes a <b>state</b>, i.e., a patch representing a limited observation of the environment, and selects an <b>action</b>, i.e., a decision that modifies the agent position and the observed state.
        <br>
          <b>(b)</b> The location of the goal is unknown during infrerence but its content can be described in various modalities:
        </p>
        <ul>
          <li><b>I:</b> aerial image patches.</li>
          <li><b>G:</b> ground-level images.</li>
          <li><b>T:</b> textual descriptions.</li>
        </ul>
          <div align="center">
            <img src="static/images/Fig_Task.png" alt="Active Geo-localization" width="90%"/>
          </div>
       </div>
      </div>
    </div>
  </div>
</section>

<!-- End paper abstract -->


<!-- Method -->
<section class="section hero">
    <div class="container is-max-desktop">
      <div class="columns is-centered has-text-centered">
        <div class="column is-four-fifths">
          <h2 class="title is-3">GeoExplorer</h2>
          <div class="content has-text-justified">

            <p>
            In this work, we introduce GeoExplorer, an AGL agent that:
            <ul>
              <li><b>1)</b> jointly predicts the <b>action-state dynamics</b></li>
              <li><b>2)</b> explores the search region with a <b>curiosity-driven reward</b></li>
            </ul>
            </p>

            <figure class="image mod-figure">
              <img src="static/images/Fig_Pipeline.png" alt="Modalities overview" width="110%">
            </figure>
            <p>
            The learning process can be divided into three stages sequentially: feature representation, Action-State Dynamics Modeling (DM), and Curiosity-Driven Exploration (CE).<br>
            <b>(a) Feature Representation.</b> The environment (st) and goal (sgoal) are encoded with different but aligned encoders, according to their modalities (e.g., aerial images (Igoal), ground-level images (Ggoal), or text (Tgoal)).<br>
            <b>(b) Action-State Dynamics Modeling.</b> A causal Transformer is trained to jointly capture action-state dynamics, guided by supervision from generated action-state trajectories for environment modeling.<br>
            <b>(c) Curiosity-Driven Exploration.</b> Based on state prediction from (b), a curiosity-driven intrinsic reward (rin) is used to encourage the agent to explore the environment by measuring the t differences between prediction and observations.
            </p>
          </div>
        </div>
      </div>
    </div>
  </section>
  <!-- End Method -->


<!-- Dataset -->
<section class="section hero is-light">
    <div class="container is-max-desktop">
      <div class="columns is-centered has-text-centered">
        <div class="column is-four-fifths">
          <h2 class="title is-3">SwissView Dataset</h2>
          <div class="content has-text-justified">

            <figure class="image mod-figure">
              <img src="static/images/Fig_Swissimage.png" alt="Modalities overview" width="60%">
            </figure>
            <p>
            Our proposed SwissView dataset is constructed from Swisstopo’s SWISSIMAGE 10cm imagery, with two distinct components:
            <ul>
              <li><b>SwissView100</b>, which comprises 100 images randomly selected from across the Swiss territory, thereby providing diverse natural and urban environment.</li>
              <li><b>SwissViewMonuments</b>, which includes 15 images of atypical or distinctive scenes, such as unusual buildings and landscapes, with corresponding ground level images.</li>
            </ul>
            </p>
          </div>
        </div>
      </div>
    </div>
  </section>
  <!-- End Dataset -->

<!-- Results -->
<section class="section hero">
  <div class="hero-body">
    <div class="container is-max-desktop">
      <div class="columns is-centered has-text-centered">
        <div class="column is-four-fifths">
          <h2 class="title is-3">Results</h2>

          <div class="content has-text-justified">
          <p>

            We evaluate GeoExplorer in four settings:
            <ul>
              <li><b>Validation</b>: we evaluate the model on the same dataset as it is trained.</li>
              <li><b>Cross-domain Transfer</b>: the model trained on the Masa dataset is evaluated on an unseen dataset using aerial view as goal modality.</li>
              <li><b>Cross-modal Generalization</b>: the goal is presented from different modalities.</li>
              <li><b>Unseen Target Generalization</b>: to evaluate the models’ adaptation to the unseen targets, we evaluate the model on our SwissView dataset, with aerial view and ground-level images.</li>
            </ul>
          </p>
        </div>

        <p><br></p>

            <h4 class="subtitle">
              <b>1. Validation and Cross-domain Generalization</b>
            </h4>

            <figure class="image mod-figure">
              <img src="static/images/WFig_cross-domain.png" alt="Modalities overview" width="60%">
            </figure>
            <p style="font-size:13px; ">
              Average success rate of GeoExplorer and the baseline over start-goal distance between 4 to 8 (C=4 to C=8) on the validation (Masa dataset) and cross-domain generalization (x-BD and Swiss-view100 datasets).
            </p>


            <p><br></p>
            <h4 class="subtitle">
              <b>2. Multimodal Generalization</b>
            </h4>

            <figure class="image mod-figure">
              <img src="static/images/WFig_cross-modal.png" alt="Modalities overview" width="60%">
            </figure>
            <p style="font-size:13px; ">
              Average success rate of GeoExplorer and the baseline over start-goal distance between 4 to 8 (C=4 to C=8) on the cross-modal generalization (MM-GAG dataset). Green, Blue and Yellow denote aerial image, ground-level image and text as the goal, respectively.
            </p>

            <p><br></p>
            <div class="finding-box mb-4" style="background-color: #e6f1f8;">
            <p class="finding-content">
              <div style="text-align: left;"><b>Takeaway I:</b> GeoExplorer with curiosity-driven intrinsic reward shows improved cross-domain and cross-modal generalization ability over baseline model with extrinsic reward only.
              </div>
            </p>
            </div>


            <p><br></p>
            <h4 class="subtitle">
              <b>3. Unseen Target Generalization</b>
            </h4>

            <figure class="image mod-figure">
              <img src="static/images/WFig_unseen.png" alt="Modalities overview" width="60%">
            </figure>
            <p style="font-size:13px; ">
              Average success rete of GeoExplorer and the baseline when C=4, C=5, and C=6 on the SwissViewMonuments dataset: (a) Aerial view as the goal; (b) Ground-level image as the goal.
            </p>


            <p><br></p>
            <div class="finding-box mb-4" style="background-color: #e6f1f8;">
            <p class="finding-content">
              <div style="text-align: left;"><b>Takeaway II:</b> GeoExplorer exhibits an impressive generalization ability in localizing unseen targets, especially when the path is long.
              </div>
            </p>
            </div>


        </div>
      </div>
    </div>
  </div>
  </section>
  <!-- End Results -->


<!-- Analysis -->
<section class="section hero">
    <div class="container is-max-desktop">
      <div class="columns is-centered has-text-centered">
        <div class="column is-four-fifths">
          <h2 class="title is-3">Does intrinsic reward improve exploration?</h2>
          <div class="content has-text-justified">
          <p>
            To provide insights of intrinsic reard and its impact on exploration, we design the following analysis and visualization:
            <ul>
              <li><b>Generated Path Visualization</b>: we visualize the generated paths of baseline and GeoExplorer, which visualize the exploration behaviour of models.</li>
              <li><b>Path Statistics</b>: we count the visited patches and end patches of all generated path, which indicates the coverage region of exploration within search area.</li>
              <li><b>Intrinsic Reward Visualization and Grounding</b>: we visualize the average intrinsic reward per patch, which provides a further understanding grounding of intrinsic reward on the exploration paths.</li>
            </ul>
          </p>
        </div>

          <p><br></p>
          <h4 class="subtitle">
              <b>1. Visualization of Exploration Ability</b>
          </h4>


          <figure class="image mod-figure">
              <img src="static/images/SFig_Unseen.png" alt="Modalities overview" width="60%">
          </figure>
          <p style="font-size:13px; ">
            <b>Generated path visualization on the SwissViewMonuments dataset.</b> Given a pair of {start (◦), goal (△)} per search area, models generate four trials with stochastic policy, randomly shown in four different colors. Compared with the baseline, the paths generated by GeoExplorer are more robust (adapted to various envrionment), diverse (different paths for the same {start, goal} pairs) and content-aware (related to state observations).
          </p>


          <p><br></p>
            <div class="finding-box mb-5" style="background-color: #e6f1f8;">
            <p class="finding-content">
              <div style="text-align: left;"><b>Takeaway III:</b> GeoExplorer shows robust, diverse, and content-related exploration ability.
              </div>
            </p>
            </div>

          <figure class="image mod-figure">
              <img src="static/images/SFig_Path.png" alt="Modalities overview" width="60%">
          </figure>
          <p style="font-size:13px; ">
             <b>Statistics of the path end and path visited on the Masa dataset.</b> (a) Statistics of the path end. We count the end location of the 895 paths in the Masa dataset test set for ground truth (goal location), GOMAA-Geo and GeoExplorer when C = 4 and C = 8. (b) Statistics of the path visited. We count all the visited patches of 895 paths in the Masa dataset test set for GOMAA-Geo and GeoExplorer when C = 4 and C = 8.
          </p>


          <p><br></p>
            <div class="finding-box mb-4" style="background-color: #e6f1f8;">
            <p class="finding-content">
              <div style="text-align: left;"><b>Takeaway IV:</b> With intrinsic reward that encourages environment exploration, GeoExplorer tends to explore more patches in the search areas, especially the patches in the center, indicating a improved exploration ability.
              </div>
            </p>
            </div>


          <p><br></p>
          <h4 class="subtitle">
              <b>2. Analysis of Intrinsic Reward</b>
          </h4>

          <figure class="image mod-figure">
              <img src="static/images/SFig_Reward.png" alt="Modalities overview" width="60%">
          </figure>
          <p style="font-size:13px; ">
             <b>Intrinsic reward visualization with images from the SwissViewMonuments dataset.</b> For each sample, from left to right: the search area, path visualization and intrinsic reward per patch. The patch with the highest intrinsic reward is highlighted with an orange rectangle in the search area. Patches with higher intrinsic reward turn out to be more “interesting”, i.e., the semantic content of these patches can hardly be predicted 546 from the surrounding patches.
          </p>


          <p><br></p>
            <div class="finding-box mb-4" style="background-color: #e6f1f8;">
            <p class="finding-content">
              <div style="text-align: left;"><b>Takeaway V:</b> Curiosity-driven intrinsic rewards provide dense, goal-agnostic, and content-related guidance to enhance the exploration ability of GeoExplorer.
              </div>
            </p>
            </div>


        </div>
      </div>
    </div>
  </section>
  <!-- End Analysis -->


<!--BibTex citation -->
  <section class="section" id="BibTeX">
    <div class="container is-max-desktop content">
      <h2 class="title">BibTeX</h2>
      <pre><code>@article{mi2025geoexplorer,
      title={{GeoExplorer}: Active Geo-localization with Curiosity-Driven Exploration},
      author={Li Mi and Manon Béchaz and Zeming Chen and Antoine Bosselut and Devis Tuia},
      year={2025},
      journal={arXiv preprint arXiv:2508.00152},
}</code></pre>
    </div>
</section>
<!--End BibTex citation -->


  <footer class="footer">
  <div class="container">
    <div class="columns is-centered">
      <div class="column is-8">
        <div class="content">

          <p>
            This page was built using the <a href="https://github.com/eliahuhorwitz/Academic-project-page-template" target="_blank">Academic Project Page Template</a> which was adopted from the <a href="https://nerfies.github.io" target="_blank">Nerfies</a> project page.
            You are free to borrow the template of this website, we just ask that you link back to this page in the footer. <br> This website is licensed under a <a rel="license"  href="http://creativecommons.org/licenses/by-sa/4.0/" target="_blank">Creative
            Commons Attribution-ShareAlike 4.0 International License</a>.
          </p>

        </div>
      </div>
    </div>
  </div>
</footer>

<!-- Statcounter tracking code -->

<!-- You can add a tracker to track page visits by creating an account at statcounter.com -->

    <!-- End of Statcounter Code -->

  </body>
  </html>