-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
144 lines (141 loc) · 8.34 KB
/
index.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
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
<div>
<link href="./index.css" rel="stylesheet" />
<script src="./index.js" type="text/javascript"></script>
<div class="home-container">
<main class="home-main">
<div class="home-hero section-container">
<div class="home-container1">
<div class="home-max-width max-content-container">
<div class="home-heading-container">
<h1 class="home-text">Towards ROXAS AI</h1>
<span class="home-text1">
Deep learning for faster and more accurate conifer cell analysis
</span>
<br>
<span class="home-text2">
by Marc Katzenmaier, Vivien Sainte Fare Garnot, Jesper
Björklund, Loïc Schneider, Jan Dirk Wegner, Georg von Arx
</span>
<br>
<div class="home-tab-selector-header">
<a
href="https://www.sciencedirect.com/science/article/pii/S1125786523000760?via%3Dihub"
target="_blank"
rel="noreferrer noopener"
class="home-link tab-selector-btn"
>
<span>Paper</span>
<br />
</a>
<a
href="https://zenodo.org/records/8369376"
class="home-link1 tab-selector-btn"
>
Dataset
</a>
<a
href="https://github.com/marckatzenmaier/TowardsRoxasAI"
class="home-link2 tab-selector-btn"
>
<span>Code</span>
<br />
</a>
</div>
</div>
</div>
</div>
<div class="home-max-width1 max-content-container">
<span class="home-text7">Abstract</span>
<p>Quantitative wood anatomy (QWA) has proven to be a powerful method for extracting relevant environmental information from tree-rings. Although classical image-analysis tools such as ROXAS have greatly improved and facilitated measurements of anatomical features, producing QWA datasets remains challenging and time-consuming. In recent years, deep learning techniques have drastically improved the performance of most computer vision tasks. We, therefore, investigate three different deep learning models (U-Net, Mask-RCNN, Panoptic Deeplab) to improve the main bottleneck, cell detection. Therefore, we create a Conifer Lumen Segmentation (CoLuS) dataset for training and evaluation. It consists of manual outlines of each cell lumen from anatomical images of several conifer species that cover a wide range of sample qualities. We furthermore apply our deep learning model to a previously published high-quality QWA chronology from Northern Finland to compare the warm-season (AMJJAS) temperature reconstruction skill of our deep learning method with that of the current ROXAS implementation, which is based on classical image analysis. Based on our evaluation dataset we show improvements of 7.6% and 8.1% for our best performing deep learning model (U-Net) for the computer vision metrics mean Intersection over Union (mIoU) and Panoptic Quality (PQ) compared to automatic ROXAS segmentation, in addition to being much faster. Furthermore, U-Net reduces the percentage error compared to automatic ROXAS analysis - which tends to systematically underestimate lumen area - by 57.8% for lumen area, 63.2% for average cell wall thickness, and 54.1% for cell count. In addition, we show higher performance for the U-Net compared to the Mask-RCNN previously used for tree cell segmentation. These improvements are independent of sample quality. For the Northern Finland QWA chronology, our U-Net model matches or outperforms ROXAS with and without manual post-processing, showing a common signal (Rbar) of 0.72 and a AMJJAS temperature correlation of 0.81 for maximum radial cell wall thickness. A clear improvement is especially visible for the anatomical latewood density, likely due to the better detection of small cell lumina. Our results demonstrate the potential of deep learning for higher-quality segmentation with lower manual post-processing time, saving weeks to months of tedious work without compromising data quality. We thus plan to implement deep learning in a future version of ROXAS.</p>
</div>
<div class="home-max-width1 max-content-container">
<span class="home-text7">Improvements over ROXAS</span>
<br/>
<img
alt="image"
src="assets/ImpovementsOverRoxas.png"
loading="lazy"
class="home-image"
/>
</div>
<div class="home-max-width2 max-content-container">
<span class="home-text8">Results</span>
<br/>
<!-- Container for the image gallery -->
<div class="container">
<!-- Full-width images with number text -->
<div class="mySlides">
<div class="numbertext">1 / 12</div>
<img src="assets/001.jpg" style="width:49%">
<img src="assets/001_.jpg" style="width:49%">
</div>
<div class="mySlides">
<div class="numbertext">2 / 12</div>
<img src="assets/002.jpg" style="width:49%">
<img src="assets/002_.jpg" style="width:49%">
</div>
<div class="mySlides">
<div class="numbertext">3 / 12</div>
<img src="assets/003.jpg" style="width:49%">
<img src="assets/003_.jpg" style="width:49%">
</div>
<div class="mySlides">
<div class="numbertext">4 / 12</div>
<img src="assets/004.jpg" style="width:49%">
<img src="assets/004_.jpg" style="width:49%">
</div>
<div class="mySlides">
<div class="numbertext">5 / 12</div>
<img src="assets/005.jpg" style="width:49%">
<img src="assets/005_.jpg" style="width:49%">
</div>
<div class="mySlides">
<div class="numbertext">6 / 12</div>
<img src="assets/006.jpg" style="width:49%">
<img src="assets/006_.jpg" style="width:49%">
</div>
<div class="mySlides">
<div class="numbertext">7 / 12</div>
<img src="assets/007.jpg" style="width:49%">
<img src="assets/007_.jpg" style="width:49%">
</div>
<div class="mySlides">
<div class="numbertext">8 / 12</div>
<img src="assets/008.jpg" style="width:49%">
<img src="assets/008_.jpg" style="width:49%">
</div>
<div class="mySlides">
<div class="numbertext">9 / 12</div>
<img src="assets/009.jpg" style="width:49%">
<img src="assets/009_.jpg" style="width:49%">
</div>
<div class="mySlides">
<div class="numbertext">10 / 12</div>
<img src="assets/010.jpg" style="width:49%">
<img src="assets/010_.jpg" style="width:49%">
</div>
<div class="mySlides">
<div class="numbertext">11 / 12</div>
<img src="assets/011.jpg" style="width:49%">
<img src="assets/011_.jpg" style="width:49%">
</div>
<div class="mySlides">
<div class="numbertext">12 / 12</div>
<img src="assets/012.jpg" style="width:49%">
<img src="assets/012_.jpg" style="width:49%">
</div>
<!-- Next and previous buttons -->
<a class="prev" onclick="plusSlides(-1)">❮</a>
<a class="next" onclick="plusSlides(1)" onload="plusSlides(0)">❯</a>
</div>
</div>
</div>
</div>
</main>
</div>
</div>
<script>
window.onload = function () {
plusSlides(0); // Delay of 5 seconds
};
</script>