Remove separator.gif

PROJ-DTRK/index.html

@@ -355,10 +355,6 @@ <h2 id="introduction">Introduction<a class="headerlink" href="#introduction" tit
 <p>The applied corrections are always the result of a complex process, involving different territory actors, until the decision is made to integrate them into the land register. In addition, land register models comes with an additional constraint linked to political decisions. Indeed, the land register models are the result of a <em>political mission</em> conducted under <em>federal laws</em> making these models of high importance and requiring constant care. We show in this research project how differences detection tool [1] of the <em>STDL 4D framework</em> can be used to emphasize and analyze these corrections along the time dimension.</p>
 <p>In addition to the constant updates of the models, changes in the reference frame can also lead to large-scale corrections of the land register models. These <em>global</em> corrections are then made even more complex by the <em>federal laws</em> that impose a high degree of correctness and accuracy.</p>
 <p>In the context of the introduction of the new reference frame <em>DM.flex</em> [2] for the Swiss land register, being able to assess the applied changes on the geographical model appear as an important aspect. Indeed, changing the reference frame for the land register models is a long and complex technical process that can be error prompt. We also show in this research project how the difference detection algorithm can be helpful to assess and verify the performed corrections.</p>
-<div align="center">
-<img src="image/separator.gif?raw=true" width="5%">
-</div>
-
 <h2 id="research-project-specifications">Research Project Specifications<a class="headerlink" href="#research-project-specifications" title="Permanent link">&para;</a></h2>
 <p>In this research project, the difference detection algorithm implemented in the <em>STDL 4D framework</em> is applied on <em>INTERLIS</em> data containing the official land register models of different Swiss <em>Canton</em>. As introduced, two main directions are considered for the difference detection algorithm :</p>
 <ul>
@@ -371,10 +367,6 @@ <h2 id="research-project-specifications">Research Project Specifications<a class
 </ul>
 <p>Through the first direction, the difference detection algorithm is presented. Considering the <em>difference models</em> it allows computing, it is shown how such model are able to extract information in between the models in order to emphasize the ability to represent, and then, to verify the evolution of the land register models.</p>
 <p>The second direction focuses on demonstrating that <em>difference models</em> are a helpful representation of the large-scale corrections that can be applied to land register during reference frame modification and how they can be used as a tool to assess the modifications and to help to fulfil the complex task of the verification of the corrected models.</p>
-<div align="center">
-<img src="image/separator.gif?raw=true" width="5%">
-</div>
-
 <h2 id="research-project-data">Research Project Data<a class="headerlink" href="#research-project-data" title="Permanent link">&para;</a></h2>
 <p>For the first research direction, the land register models of the <em>Thurgau Kanton</em> are considered. They are selected in order to have a small temporal distance allowing to focus on a small amount of well-defined differences :</p>
 <ul>
@@ -400,10 +392,6 @@ <h2 id="research-project-data">Research Project Data<a class="headerlink" href="
 <p><em>Canton of Geneva</em>, <strong>2019-04</strong>, <em>INTERLIS</em></p>
 </li>
 </ul>
-<div align="center">
-<img src="image/separator.gif?raw=true" width="5%">
-</div>
-
 <h2 id="difference-models-a-temporal-derivative">Difference Models : A Temporal Derivative<a class="headerlink" href="#difference-models-a-temporal-derivative" title="Permanent link">&para;</a></h2>
 <p>This first section focuses on short-term differences to show how difference models work and how they are able to represent the modifications extracted out of the two compared models. The following images give an illustration of the considered dataset, which are the land register models of <em>Thurgau Kanton</em> :</p>
 <div align="center" style="font-style: italic">
@@ -454,10 +442,6 @@ <h2 id="difference-models-a-temporal-derivative">Difference Models : A Temporal
 </div>
 
 <p>One can see how <em>difference models</em> can be used to track down modifications brought to the land register in a simple manner, while keeping the information of the unchanged elements between the two compared models. This demonstrates that information that exists between models can be extracted and represented for further users or automated processes. In addition, such <em>difference models</em> can be computed at any scale, considering small area up to the whole countries.</p>
-<div align="center">
-<img src="image/separator.gif?raw=true" width="5%">
-</div>
-
 <h2 id="difference-models-an-assessment-tool">Difference Models : An Assessment Tool<a class="headerlink" href="#difference-models-an-assessment-tool" title="Permanent link">&para;</a></h2>
 <p>On the previous section, the <em>difference models</em> are computed using two models only separated of a few days, containing only a small amount of clear and simple modifications. This section focuses on detecting differences on larger models, separated by several years. In this case, the land register of the <em>Canton of Geneva</em> is considered :</p>
 <div align="center" style="font-style: italic">

PROJ-GEPOOL/index.html

@@ -340,10 +340,6 @@ <h2 id="introduction">Introduction<a class="headerlink" href="#introduction" tit
 <p>The Canton of Geneva manages a register of swimming pools, counting - in principle - all and only those swimming pools that are in-ground or, at least, permanently fixed to the ground. The swimming pool register is part of a far more general cadastre, including several other classes of objects (cf. <a href="https://ge.ch/sitg/fiche/3205">this page</a>). </p>
 <p>Typically the swimming pool register is updated either by taking building/demolition permits into account, or by manually checking its multiple records (4000+ to date) against aerial images, which is quite a long and tedious task. Exploring the opportunity of leveraging Machine Learning to help domain experts in such an otherwise tedious tasks was one of the main motivations behind this study. As such, no prior requirements/expectations were set by the recipients. </p>
 <p>The study was autonomously conducted by the STDL team, using Open Source software and Open Data published by the Canton of Geneva. Domain experts were asked for feedback only at a later stage. In the following, details are provided regarding the various steps we followed. We refer the reader to <a href="../TASK-IDET/">this page</a> for a thorough description of the generic STDL Object Detection Framework.</p>
-<div align="center">
-<img src="image/separator.gif?raw=true" width="5%">
-</div>
-
 <h2 id="method">Method<a class="headerlink" href="#method" title="Permanent link">&para;</a></h2>
 <p>Several steps are required to set the stage for object detection and eventually reach the goal of obtaining - ideally - even more than decent results. Despite the linear presentation that the reader will find here-below, multiple back-and-forths are actually required, especially through steps 2-4.</p>
 <h3 id="1-data-preparation">1. Data preparation<a class="headerlink" href="#1-data-preparation" title="Permanent link">&para;</a></h3>
@@ -442,10 +438,6 @@ <h3 id="4-prediction-assessment">4. Prediction assessment<a class="headerlink" h
 
 <p>The latter figure was obtained by evaluating the predictions of our best model over the test dataset. Inferior models exhibited a similar behavior, with a downward offset in terms of <span class="arithmatex">\(F_1\)</span> score. In practice, upon <strong>iterating over multiple realizations</strong> (with different hyper-parameters, training data and so on) we aimed at maximizing the value of the <span class="arithmatex">\(F_1\)</span> score on the validation dataset, and stopped when the <span class="arithmatex">\(F_1\)</span> score went over the value of 90%.</p>
 <p>As the ground-truth data we used turned out not to be 100% accurate, the responsibility for mismatching predictions has to be shared between ground-truth data and the predictive model, at least in some cases. In a more ideal setting, ground-truth data would be 100% accurate and differences between a given metric (precision, recall, <span class="arithmatex">\(F_1\)</span> score) and 100% should be imputed to the model.</p>
-<div align="center">
-<img src="image/separator.gif?raw=true" width="5%">
-</div>
-
 <h2 id="domain-experts-feedback">Domain experts feedback<a class="headerlink" href="#domain-experts-feedback" title="Permanent link">&para;</a></h2>
 <p>All the predictions having a score <span class="arithmatex">\(\geq\)</span> 5% obtained by our best model were exported to Shapefile and shared with the experts in charge of the cadastre of the Canton of Geneva, who carried out a thorough evaluation. By checking predictions against the swimming pool register as well as aerial images, it was empirically found that the threshold on the minimum score (= <code>thr</code>) should be set as high as 97%, in order not to have too many false positives to deal with. In spite of such a high threshold, 562 potentially new objects were detected (over 4652 objects which were known when this study started), of which:</p>
 <ul>
@@ -472,10 +464,6 @@ <h2 id="domain-experts-feedback">Domain experts feedback<a class="headerlink" hr
 <i>Examples of detected swimming pools which are not subject to registration: placed on top of a building (left), inflatable hence temporary (right).</i>
 </p>
 
-<div align="center">
-<img src="image/separator.gif?raw=true" width="5%">
-</div>
-
 <h2 id="conclusion">Conclusion<a class="headerlink" href="#conclusion" title="Permanent link">&para;</a></h2>
 <p>The analysis reported in this document confirms the opportunity of using state-of-the-art Deep Learning approaches to assist experts in some of their tasks, in this case that of keeping the cadastre up to date. Not only the opportunity was explored and actually confirmed, but valuable results were also produced, leading to the detection of previously unknown objects. At the same time, our study also shows how essential domain expertise still remains, despite the usage of such advanced methods.</p>
 <p>As a concluding remark, let us note that our predictive model may be further improved. In particular, it may be rendered less prone to false positives, for instance by:</p>

PROJ-HETRES/index.html

@@ -607,9 +607,11 @@ <h3 id="42-image-processing">4.2 Image processing<a class="headerlink" href="#42
 
 <h4 id="421-statistical-tests-on-the-original-and-filtered-pixels">4.2.1 Statistical tests on the original and filtered pixels<a class="headerlink" href="#421-statistical-tests-on-the-original-and-filtered-pixels" title="Permanent link">&para;</a></h4>
 <p>The statistical tests were performed on the original and filtered pixels.</p>
-<p>Two low pass filters were tested:
-- Gaussian with a sigma of 5;
-- Bilinear downsampling with scale factors of 1/3, 1/5 and 1/17, corresponding to resolutions of 9, 15 and 50 cm.</p>
+<p>Two low pass filters were tested:</p>
+<ul>
+<li>Gaussian with a sigma of 5;</li>
+<li>Bilinear downsampling with scale factors of 1/3, 1/5 and 1/17, corresponding to resolutions of 9, 15 and 50 cm.</li>
+</ul>
 <p>In the original and the filtered cases, the pixels for each GT tree were extracted from the images and sorted by class. Then, the corresponding NDVI is computed. Each pixel has 5 attributes corresponding to its value on the four bands (R, G, B, NIR) and its NDVI.<br />
 First, the per-class boxplots of the attributes were executed to see if the distinction between classes was possible on one or several bands or on the NDVI.<br />
 Then, the principal component analysis (PCA) was computed on the same values to see if their linear combination allowed the distinction of the classes.</p>
@@ -840,7 +842,7 @@ <h4 id="532-vegetation-healthy-index">5.3.2 Vegetation Healthy Index<a class="he
 <li>unhealthy and other trees in 2021</li>
 <li>dead and unhealthy trees in 2020 and 2022</li>
 </ul>
-<p>Explanations similar to those for NDVI may partly explain the significance obtained. In any case,it is encouraging that the VHI helps to differentiate health classes thanks to different evolution through the years. </p>
+<p>Explanations similar to those for NDVI may partly explain the significance obtained. In any case,it is encouraging that the VHI helps to differentiate health classes thanks to different evolution through the years.</p>
 <h3 id="54-random-forest">5.4 Random Forest<a class="headerlink" href="#54-random-forest" title="Permanent link">&para;</a></h3>
 <p>The results of the RF that are presented and discussed are: (1) the optimization and ablation study, (2) the ground truth analysis, (3) the predictions for the AOI and (4) the performance with downgraded data. </p>
 <h4 id="541-optimization-and-ablation-study">5.4.1 Optimization and ablation study<a class="headerlink" href="#541-optimization-and-ablation-study" title="Permanent link">&para;</a></h4>