Skip to content

Commit 22421a6

Browse files
committed
update contents
1 parent fc072cc commit 22421a6

File tree

4 files changed

+41
-36
lines changed

4 files changed

+41
-36
lines changed

docs/index.html

Lines changed: 39 additions & 35 deletions
Original file line numberDiff line numberDiff line change
@@ -3,10 +3,10 @@
33
<head>
44
<meta charset="utf-8">
55
<meta name="description"
6-
content="Enhanced Camera-Radar Object Detection with Cross-modality Knowledge Distillation.">
7-
<meta name="keywords" content="CRKD">
6+
content="MemFusionMap: Working Memory Fusion for Online Vectorized HD Map Construction">
7+
<meta name="keywords" content="MemFusionMap">
88
<meta name="viewport" content="width=device-width, initial-scale=1">
9-
<title>CRKD: Enhanced Camera-Radar Object Detection with Cross-modality Knowledge Distillation</title>
9+
<title>MemFusionMap: Working Memory Fusion for Online Vectorized HD Map Construction</title>
1010

1111
<!-- Global site tag (gtag.js) - Google Analytics -->
1212
<script async src="https://www.googletagmanager.com/gtag/js?id=G-PYVRSFMDRL"></script>
@@ -91,9 +91,13 @@
9191
<h1 class="title is-1 publication-title">MemFusionMap: Working Memory Fusion for Online Vectorized HD Map Construction</h1>
9292
<div class="is-size-5 publication-authors">
9393
<span class="author-block">
94-
<a href="https://junshao0104.github.io/">Lingjun Zhao</a><sup>1*</sup>,</span>
94+
<a href="https://song-jingyu.github.io/">Jingyu Song</a><sup>1</sup>,</span>
9595
<span class="author-block">
96-
<a href="https://www.linkedin.com/in/jingyu-song-93763a132/">Jingyu Song</a><sup>1*✉</sup>,</span>
96+
<a href="">Xudong Chen</a><sup>2</sup>,</span>
97+
<span class="author-block">
98+
<a href="">Liupei Lu</a><sup>2</sup>,</span>
99+
<span class="author-block">
100+
<a href="">Jie Li</a><sup>2</sup>,</span>
97101
<span class="author-block">
98102
<a href="https://fieldrobotics.engin.umich.edu/">Katherine A. Skinner</a><sup>1</sup>
99103
</span>
@@ -112,16 +116,17 @@ <h1 class="title is-1 publication-title">MemFusionMap: Working Memory Fusion for
112116
</div>
113117

114118
<div class="is-size-5 publication-authors">
115-
<span class="author-block"><sup>*</sup>Equal Contribution</span>
119+
<!-- <span class="author-block"><sup>*</sup>Equal Contribution</span> -->
116120
<!-- <span class="author-block"><sup>†</sup>Corresponding Author</span> -->
117121
<span class="author-block"><sup>1</sup>University of Michigan</span>
122+
<span class="author-block"><sup>2</sup>NVIDIA</span>
118123
</div>
119124

120125
<div class="column has-text-centered">
121126
<div class="publication-links">
122127
<!-- PDF Link. -->
123128
<span class="link-block">
124-
<a href="https://arxiv.org/pdf/2403.19104.pdf"
129+
<a href="https://arxiv.org/pdf/2409.18737"
125130
class="external-link button is-normal is-rounded is-dark">
126131
<span class="icon">
127132
<i class="fas fa-file-pdf"></i>
@@ -130,7 +135,7 @@ <h1 class="title is-1 publication-title">MemFusionMap: Working Memory Fusion for
130135
</a>
131136
</span>
132137
<span class="link-block">
133-
<a href="https://arxiv.org/abs/2403.19104"
138+
<a href="https://arxiv.org/abs/2409.18737"
134139
class="external-link button is-normal is-rounded is-dark">
135140
<span class="icon">
136141
<i class="ai ai-arxiv"></i>
@@ -139,23 +144,23 @@ <h1 class="title is-1 publication-title">MemFusionMap: Working Memory Fusion for
139144
</a>
140145
</span>
141146
<!-- Video Link. -->
142-
<span class="link-block">
147+
<!-- <span class="link-block">
143148
<a href="https://youtu.be/A63ho5ESz8w?si=i_xoSdJHlDclfyUF"
144149
class="external-link button is-normal is-rounded is-dark">
145150
<span class="icon">
146151
<i class="fab fa-youtube"></i>
147152
</span>
148153
<span>Video</span>
149154
</a>
150-
</span>
155+
</span> -->
151156
<!-- Code Link. -->
152157
<span class="link-block">
153-
<a href="https://github.com/Song-Jingyu/CRKD"
158+
<a href="https://github.com/Song-Jingyu/MemFusionMap"
154159
class="external-link button is-normal is-rounded is-dark">
155160
<span class="icon">
156161
<i class="fab fa-github"></i>
157162
</span>
158-
<span>Code</span>
163+
<span>Code (Coming soon)</span>
159164
</a>
160165
</span>
161166
<!-- Dataset Link. -->
@@ -186,7 +191,7 @@ <h1 class="title is-1 publication-title">MemFusionMap: Working Memory Fusion for
186191
<!-- add image docs/static/images/pitch.svg-->
187192
<img src="./static/images/pitch.svg" alt="CRKD" style="width: 100%; height: auto;">
188193
<h2 class="subtitle has-text-centered">
189-
We propose CRKD to conduct a novel cross-modality knowledge distillation path from LiDAR-camera teacher to camera-radar student.
194+
We proposed MemFusionMap, a novel approach for effective online vectorized HD map construction with enhanced temporal reasoning capability.
190195
</h2>
191196
</div>
192197
</div>
@@ -258,13 +263,11 @@ <h2 class="subtitle has-text-centered">
258263
<div class="column is-four-fifths">
259264
<h2 class="title is-3">Abstract</h2>
260265
<div class="content has-text-justified">
261-
In the field of 3D object detection for autonomous driving, LiDAR-Camera (LC) fusion is the top-performing sensor configuration.
262-
Still, LiDAR is relatively high cost, which hinders adoption of this technology for consumer automobiles.
263-
Alternatively, camera and radar are commonly deployed on vehicles already on the road today, but performance of Camera-Radar (CR) fusion falls behind LC fusion.
264-
In this work, we propose CRKD to bridge the performance gap between LC and CR detectors with a novel cross-modality knowledge distillation (KD) framework.
265-
We use the Bird's-Eye-View (BEV) representation as the shared feature space to enable effective knowledge distillation.
266-
To accommodate the unique cross-modality KD path, we propose four distillation losses to help the student learn crucial features from the teacher model.
267-
We present extensive evaluations on the nuScenes dataset to demonstrate the effectiveness of the proposed CRKD framework.
266+
High-definition (HD) maps provide environmental information for autonomous driving systems and are essential for safe planning.
267+
While existing methods with single-frame input achieve impressive performance for online vectorized HD map construction, they still struggle with complex scenarios and occlusions.
268+
We propose MemFusionMap, a novel temporal fusion model with enhanced temporal reasoning capabilities for online HD map construction.
269+
Specifically, we contribute a working memory fusion module that improves the model's memory capacity to reason across a history of frames. We also design a novel temporal overlap heatmap to explicitly inform the model about the temporal overlap information and vehicle trajectory in the Bird's Eye View space. By integrating these two designs, MemFusionMap significantly outperforms existing methods while also maintaining a versatile design for scalability.
270+
We conduct extensive evaluation on open-source benchmarks and demonstrate a maximum improvement of 5.4% in mAP over state-of-the-art methods.
268271
</div>
269272
</div>
270273
</div>
@@ -284,6 +287,10 @@ <h2 class="title is-3">Video</h2>
284287
</div>
285288
</section>
286289

290+
\item
291+
\item We propose a novel design of maintaining a temporal overlap heatmap, providing a strong cue for the model to reason across a history of frames and also implicitly encoding valuable insights of the vehicle's trajectory.
292+
\item
293+
287294
<!-- Overview section -->
288295
<section class="section">
289296
<div class="container is-max-desktop">
@@ -292,23 +299,20 @@ <h2 class="title is-3">Video</h2>
292299
<h2 class="title is-3">Overview</h2>
293300
<div class="content has-text-justified">
294301
<p>
295-
We propose a novel cross-modality KD framework to enable LC-to-CR distillation in the BEV feature space. With the transferred knowledge
296-
from an LC teacher detector, the CR student detector can outperform existing baselines without additional cost during inference.
302+
We propose a simple yet effective model to fuse working memory features in BEV space for online vectorized HD map construction.
303+
MemFusionMap focuses on improving the network's temporal reasoning capability while also maintaining a versatile design for scalability and compatibility.
297304
</p>
298305
<p>
299-
We design four KD modules to address the notable discrepancies between different sensors to realize realize effective cross-modality KD.
300-
As we operate KD in the BEV space, the proposed loss designs can be applied to other KD configurations.
301-
Our improvement also includes adding a gated network to the baseline model for adaptive fusion.
306+
We propose a novel design of maintaining a temporal overlap heatmap, providing a strong cue for the model to reason across a history of frames and also implicitly encoding valuable insights of the vehicle's trajectory.
302307
</p>
303308
<p>
304-
We conduct extensive evaluation on nuScenes to demonstrate the effectiveness of CRKD.
305-
CRKD can improve the mAP and NDS of student detectors by 3.5% and 3.2% respectively.
306-
Since our method focuses on a novel KD path with distinctively large modality gap, we provide thorough study and analysis to support our design choices.
309+
We conduct extensive evaluation on nuScenes and Argoverse2 to demonstrate the effectiveness of MemFusionMap.
310+
The proposed method significantly outperforms the state-of-the-art method, achieving a maximum improvement of 5.4% in mAP.
307311
</div>
308312
</div>
309313
</div>
310314
<!-- add an image docs/static/images/Overall_Diagram.svg -->
311-
<img src="./static/images/Overall_Diagram.jpg" alt="CRKD Overview" style="width: 100%; height: auto;">
315+
<img src="./static/images/Overall_Diagram.svg" alt="MemFusionMap Overview" style="width: 100%; height: auto;">
312316
</div>
313317
</section>
314318

@@ -319,12 +323,12 @@ <h2 class="title is-3">Overview</h2>
319323
<section class="section" id="BibTeX">
320324
<div class="container is-max-desktop content">
321325
<h2 class="title">BibTeX</h2>
322-
<pre><code>@inproceedings{zhao2024crkd,
323-
author = {Zhao, Lingjun and Song, Jingyu and Skinner, Katherine A},
324-
title = {CRKD: Enhanced Camera-Radar Object Detection with Cross-modality Knowledge Distillation},
325-
journal = {2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
326-
year = {2024},
327-
}</code></pre>
326+
<pre><code>@article{song2024memfusionmap,
327+
title={MemFusionMap: Working Memory Fusion for Online Vectorized HD Map Construction},
328+
author={Song, Jingyu and Chen, Xudong and Lu, Liupei and Li, Jie and Skinner, Katherine A},
329+
journal={arXiv preprint arXiv:2409.18737},
330+
year={2024}
331+
}</code></pre>
328332
</div>
329333
</section>
330334

-208 KB
Binary file not shown.

docs/static/images/Overall_Diagram.svg

Lines changed: 1 addition & 0 deletions
Loading

docs/static/images/pitch.svg

Lines changed: 1 addition & 1 deletion
Loading

0 commit comments

Comments
 (0)