-
Notifications
You must be signed in to change notification settings - Fork 0
/
overview.html
73 lines (59 loc) · 5.47 KB
/
overview.html
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
<!doctype html>
<html lang="en" style="scroll-behavior: smooth;">
<head>
<title>All-In-One Drive</title>
<!-- Required meta tags -->
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no">
<link href="main.css" rel="stylesheet">
<!-- Bootstrap CSS -->
<link rel="stylesheet" href="https://stackpath.bootstrapcdn.com/bootstrap/4.3.1/css/bootstrap.min.css"
integrity="sha384-ggOyR0iXCbMQv3Xipma34MD+dH/1fQ784/j6cY/iJTQUOhcWr7x9JvoRxT2MZw1T" crossorigin="anonymous">
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/5.9.0/css/all.min.css">
</head>
<body>
<header>
<nav id="navbar"></nav>
</header>
<main role="main">
<div class="container">
<div class="row">
<div class="col-md-12">
<h1 class="about-group-name">Overview</h1>
<!--introduction and motivation-->
<p><img src="http://www.xinshuoweng.com/papers/AIODrive/main.png" style="width:41%; float:right;">The All-in-One Drive (AIODrive) is a large-scale comprehensive perception dataset for autonomous driving developed by Carnegie Mellon University in order to innovate multi-sensor multi-task perception systems for autonomous driving. Though various driving datasets have been released to advance perception systems, these datasets often have different objectives and provide different annotations, sensors and environmental variations such as 3D segmentation (SemanticKITTI), rich maps (Argoverse), radar sensing (nuScenes), high-speed driving (A*3D), adverse weather and large-scale data (Waymo). In contrast, our goal is to build a comprehensive dataset that combines the strengths of all existing perception datasets, in order to further innovate multi-sensor multi-task perception systems. To that end, our AIODrive dataset provides the most comprehensive sensing modalities, support annotations for all mainstream perception tasks and collect diverse driving environmental variations. Additionally, one unique feature of our AIODrive dataset is that we provide various point cloud data such as point cloud generated from SPAD-LiDAR and point cloud with different density levels (beyond the density of normal Velodyne-64 LiDAR), which we hope can encourage research investigating how much the point cloud density will affect the performance of perception systems.</p>
<!--technical details-->
<p>One fact we realized is that building a real-world large-scale dataset that forms a union of the various strengths of currently available datasets is challenging as it requires significant resources, e.g., several years of effort and billions of dollars. As a result, our current solution is the use of a simulator, CARLA, to generate our comprehensive perception dataset. Synthetic data generation is able to meet the challenges of creating a comprehensive perception dataset because: (1) Diverse data can be generated in simulation as the Carla simulator can change the density of traffic, adjust distribution of the agent motion, generate violations of traffic rules and change weather and lighting conditions; (2) Large amounts of annotated data for a multitude of tasks can be generated by combining and post-processing standard Carla outputs. For example, we can easily project the 2D semantic annotation to 3D given the corresponding depth image, resulting in the 3D semantic annotation for point clouds. (3) A ‘physical’ sensing platform can easily be constructed in simulation without costing millions of dollars, and we can change sensor configuration and even create new sensors (e.g., long-range high-density LiDAR and SPAD-LiDAR) that not currently available in the real world.</p>
<!--licensing-->
<p>Our AIODrive dataset is free to use for non-commercial purposes. Examples of non-commercial use include but are not limited to personal use, educational use, and research use. If you intend to use the AIODrive dataset for commercial purposes, please contact us for commercial licensing options by sending an e-mail to [email protected].</p>
<!--citation-->
<p>Please cite our publications <a href="publications.html">here</a> when using our AIODrive dataset.
</p>
</div>
</div>
<hr class="featurette-divider">
</div>
</main>
<!-- FOOTER -->
<footer id="footer"></footer>
<!-- Optional JavaScript -->
<!-- jQuery first, then Popper.js, then Bootstrap JS -->
<script src="https://code.jquery.com/jquery-3.4.1.min.js"
integrity="sha256-CSXorXvZcTkaix6Yvo6HppcZGetbYMGWSFlBw8HfCJo=" crossorigin="anonymous"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/1.14.7/umd/popper.min.js"
integrity="sha384-UO2eT0CpHqdSJQ6hJty5KVphtPhzWj9WO1clHTMGa3JDZwrnQq4sF86dIHNDz0W1" crossorigin="anonymous">
</script>
<script src="https://stackpath.bootstrapcdn.com/bootstrap/4.3.1/js/bootstrap.min.js"
integrity="sha384-JjSmVgyd0p3pXB1rRibZUAYoIIy6OrQ6VrjIEaFf/nJGzIxFDsf4x0xIM+B07jRM" crossorigin="anonymous">
</script>
<script src="prism.js"></script>
<script src="prism-bibtex.js"></script>
<script src="main.js"></script>
<script>
$(document).ready(function () {
make_navbar("AIODrive");
insert_footer();
});
</script>
</body>
</html>