Updated webpage

index.md

@@ -1,34 +1,36 @@
-<!-- Airborne particles such as dust, smoke and fog have a significant detrimental impact on Lidar-based robotic perception systems. Lidar rays can reflect on these particles, leading modern perception methods to erroneous results, such as false obstacles or misclassified elements. We propose a method to detect airborne particles in 3D Lidar point clouds using classification from geometric features and Lidar intensity returns. We compare three different classifiers and we evaluate our approach using real dust and fog data collected in outdoor scenarios. We achieve an accuracy of up to 95% in detecting airborne particles in Lidar point clouds, making our proposed method a promising solution for applications such as obstacle detection and object recognition in outdoor environments.  -->
-Final abstract goes here
+<h2>Abstract</h2>
+LiDAR sensors have been very popular in robotics due to their ability to provide accurate range measurements and their robustness to lighting conditions. However,  their sensitivity to airborne particles such as dust or fog can lead to perception algorithm failures (e.g. the detection of false obstacles by field robots). In this work, we address this problem by proposing methods to classify airborne particles in LiDAR data. We propose and compare two deep learning approaches, the first is based on voxel-wise classification, while the second is based on point-wise classification. We also study the impact of different combinations of input features extracted from LiDAR data, including the use of multi-echo returns as a classification feature. We evaluate the performance of the proposed methods on a realistic dataset with the presence of fog and dust particles in outdoor scenes. We achieve an F1 score of 94\% for the classification of airborne particles in LiDAR point clouds, thereby significantly outperforming the state-of-the-art. We show the practical significance of this work on two real-world use cases: a relative pose estimation task using point cloud matching, and an obstacle detection task.
 <br/>
 <br/>
 
-<!-- <center>
-<img src="imgs/pcl_seg.png" alt="DustSegmentation" width="500px"/>
+<center>
+<img src="imgs/intro.png" alt="DustSegmentation" width="500px"/>
 <br/>
-Image showing the successful detection and segmentation of Lidar points generated by dust.
-<br/> -->
+Left: Image of an experimental scene with dust behind a car. Middle: LiDAR point cloud with dust corruption (colored by height). Right: Predicted particles (white) and non-particles (red)
+<br/> 
+</center>
 
-<h2>Architecture Comparison</h2>
-
-<h3>Voxel Classification Approach</h3>
-
-<h3>Lidar Image Classification Approach</h3>
-
-<h2>Quantitative Results</h2>
-
-<!-- <img src="imgs/results.png" alt="Results" width="400px"/> -->
-
-<h2>Qualitative Results</h2>
-
-<!-- <img src="imgs/fog_pred.png" alt="FogPrediction" width="300px"/>
-
-<img src="imgs/dust_pred.png" alt="DustPrediction" width="300px"/>
-
-<img src="imgs/dust_natural_pred.png" alt="DustNaturalPrediction" width="300px"/> -->
+<h2> Classification Performance </h2>
+<center>
+<img src="imgs/quantitative_res.png" alt="results" width="500px"/>
+<br/>
+Comparison of classification results between our approaches with the best configurations of input features and the state of the art.
+<br/> 
+</center>
 
 <h2> Used Cases </h2>
 
 <h3>Point Cloud Matching Task</h3>
 
 <h3>Obstacle Detection Task</h3>
+<center>
+<img src="imgs/og_final.png" alt="og" width="500px"/>
+<br/>
+Occupancy grids (occupied: black cells, free: light grey cells, unobserved: dark grey) and predicted point cloud with particles in white and non-particlesin red. Left:Original point cloud. Right: Point cloud with predicted particles removed.
+<br/> 
+
+<img src="imgs/og_res.png" alt="og" width="500px"/>
+<br/>
+Performance of our classification approaches on the obstacle detection task
+<br/> 
+</center>