Kevin/website
1
0
Fork 0
mirror of https://github.com/LJ5O/website.git synced 2026-08-11 22:54:46 +02:00
Code Issues Packages Projects Releases Wiki Activity

Continued to write about hachinohe internship

Browse Source
This commit is contained in:
LJ5O
2025-04-28 16:55:12 +02:00
parent ba2c95f370
commit 77a384f647
4 changed files with 83 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/assets/base.css
View File

@@ -1,7 +1,7 @@
* {
/** {
overflow-y: hidden;
overflow-x: hidden;
}
}*/
html{
background-color: #1E1E1E;
}

18
src/assets/localization/en/localization.json
View File

@@ -116,7 +116,23 @@
"content": {
"context": {
"title": "Some context before begining...",
"text": "The picture you can see next to this text is Hachinohe Kosen, the place were I had the chance to go for my very first research internship. I was sent there for a 3 months stay, with some other French students."
"text": "The picture you can see next to this text is Hachinohe Kosen, the place where I had the chance to go for my very first research internship. I was sent there for a 3 months stay, with some other French students."
},
"semantic": {
"title": "Semantic Segmentation models study",
"text": "I installed three models on my laptop, Unet, Segnet and DeeplabV3, and trained them using the \"oxford_iiit_pet\" dataset ( from which the pictures on this page come from ). I searched about them and obtained various statistics and results, that I analysed and used to determine which one was the best Semantic Segmentation model back then.",
"table": {
"model": "Model",
"accuracy": "Accuracy",
"loss": "Loss",
"process_time": "Processing time for one frame",
"year": "Release year"
},
"result": "According to this table, the best model among those I tested was DeeplabV3. It was the one giving the best results according to Accuracy, but is a little slower than Unet."
},
"human_recognition": {
"title": "Human detection",
"text": "Along with the Semantic Segmentation, I was asked to find a way to recognize an human from a webcam stream. I chose the model YoloV5 to implement this functionality, and got some nice results."
}
}
}

2
src/assets/main.css
View File

@@ -1,6 +1,6 @@
@import './base.css';
h1,h2,h3,h4,h5,p{
h1,h2,h3,h4,h5,p,td,th{
color: #EAEAEA;
/*Family Roboto or Atma ?*/
}

63
src/views/projects/semanticSegmentationView.vue
View File

@@ -11,6 +11,49 @@
<p>{{ $t("projects.semantic_segmentation.content.context.text") }}</p>
<img :src="hachinoheKosenImg"/>
</doubleRowHolder>
<h2>{{ $t("projects.semantic_segmentation.content.semantic.title") }}</h2>
<p>{{ $t("projects.semantic_segmentation.content.semantic.text") }}</p>
<div id="table_wrapper">
<table>
<thead>
<tr>
<th>{{ $t("projects.semantic_segmentation.content.semantic.table.model") }}</th>
<th>{{ $t("projects.semantic_segmentation.content.semantic.table.accuracy") }}</th>
<th>{{ $t("projects.semantic_segmentation.content.semantic.table.loss") }}</th>
<th>{{ $t("projects.semantic_segmentation.content.semantic.table.process_time") }}</th>
<th>{{ $t("projects.semantic_segmentation.content.semantic.table.year") }}</th>
</tr>
</thead>
<tbody>
<tr>
<td>Unet</td>
<td>0.9233</td>
<td>0.1898</td>
<td>10 ms</td>
<td>2015</td>
</tr>
<tr>
<td>Segnet</td>
<td>0.9681</td>
<td>0.07575</td>
<td>100 ms</td>
<td>2016</td>
</tr>
<tr>
<td>DeeplabV3</td>
<td>0.9708</td>
<td>0.06759</td>
<td>15 ms</td>
<td>2017</td>
</tr>
</tbody>
</table>
</div>
<p>{{ $t("projects.semantic_segmentation.content.semantic.result") }}</p>
<h2>{{ $t("projects.semantic_segmentation.content.human_recognition.title") }}</h2>
<p>{{ $t("projects.semantic_segmentation.content.human_recognition.text") }}</p>
</article>
</template>
@@ -50,5 +93,25 @@
min-width: 200px;
max-width: 400px;
}
#table_wrapper{
margin-top: 20px;
margin-bottom: 15px;
width: 80%;
min-width: 250px;
overflow-x: auto !important;
}
th{
font-size:1.3em;
padding-left: 5px;
padding-right: 5px;
}
table,th,td{
border: solid 1px white;
border-collapse: collapse;
}
table{
margin: auto;
}
</style>