-
Notifications
You must be signed in to change notification settings - Fork 0
/
ToyotaDesign2023.html
76 lines (70 loc) · 2.78 KB
/
ToyotaDesign2023.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
<!DOCTYPE html>
<html>
<head>
<link rel="stylesheet" href="Style.css" />
<title>Miekale Smith's Portfoilo</title>
<div class="nav">
<ol>
<li><a href="Home.html#home">HOME</a></li>
<li><a href="Home.html#projects">PROJECTS</a></li>
<li><a href="Home.html#connect">CONNECT</a></li>
</ol>
</div>
</head>
<body style="background-color: #2a2c36">
<div class="project-pages">
<h1>Toyota Design Challenge 2023</h1>
<div class="slideshow">
<div class="slide">
<img src="media/Toyota_sticker_hole_detection.jpg" alt="" />
</div>
<div class="slide">
<img src="media/Toyota_results.png" alt="" />
</div>
<a class="prev" onclick="plusSlides(-1)">❮</a>
<a class="next" onclick="plusSlides(1)">❯</a>
<div style="text-align: center">
<span class="dot" onclick="currentSlide(1)"></span>
<span class="dot" onclick="currentSlide(2)"></span>
</div>
</div>
<div class="project-pages-main">
<ul>
<li>
Developed an object detection model using pretrained YOLO V8 to
distinguish between three similar object classes.
</li>
<li>
Utilized a dataset labeled with Roboflow for the 2023 Toyota design
challenge, involving detecting stickers on holes under various
conditions.
</li>
</ul>
<p>
The 2023 Toyota design challenge required competitors to detect
whether a hole was covered by a sticker, partially covered, or whether there
was no sticker on the hole. To make the problem challenging, holes and
stickers were similar sizes and detection methods had to detect the three
classes with dynamic camera positioning and varied background color.
</p>
<p>
Addressing the challenge, we created a labeled dataset of 400+ images,
implementing transfer learning with YOLO v8. Semi-supervised learning
techniques were applied by using model predictions for training. The model
achieved <b>82%</b> accuracy on stickers, <b>89%</b> on holes, and <b>92%</b> on partially
covered holes. It demonstrated recognition across dynamic camera
positions on red, white, and metallic surfaces.
</p>
<div class="icon" style="text-align: center">
<a href="https://github.com/hepromark/ToyotaChallenge2023"
><img src="media/Github_symbol.png" alt="JestureBot Github"
/></a>
</div>
</div>
<div class="back-button">
<a href="Home.html#projects">BACK TO PROJECTS</a>
</div>
</div>
<script src="scripts.js"></script>
</body>
</html>