Tag: deep learning
-
2D Boxes to 3D Frustums: Simplifying Point Cloud Labeling for Object Detection
Labeling point clouds for object detection is a daunting task. The data is massive and its 3D nature makes it harder for the annotator to find the objects of interest as compared to labeling images. In order to facilitate the process of labeling point clouds, I have explored and developed a simple algorithm that leverages…
-
Synchronizing LiDAR and Camera Data for Offline Processing Using ROS
Robots and other complex machines are usually equipped with a wide variety of sensors. Often, when performing tasks such as driving autonomously or collecting data for offline processing, it is useful to synchronize the different sensors. For example, in my work with 3D Machine Learning, I had to collect a dataset consisting of LiDAR point…
-
3D Object Detection with Open3D-ML and PyTorch Backend
In previous articles, I described how I used Open3D-ML to do Semantic Segmentation on the SemanticKITTI dataset and on my own dataset. Now it is time to move to another important aspect of the Perception Stack for Autonomous Vehicles and Robots, which is Object Detection from Point Clouds. Make sure to install Open3D-ML with PyTorch…
-
Filtering a Point Cloud to Match the Field of View of the Camera
In a previous post, I described why and how I was collecting a Point Clouds dataset. My setup is depicted in the image above, where a 360°, 32-beam LiDAR is placed above a stereo camera. One of the steps mentioned in the article was to crop (or filter) the Point Cloud to only show points approximately…
-
Installing Open3D-ML for 3D Computer Vision with PyTorch
In a previous post, I introduced my reasons to test Open3D-ML and the steps to install it with TensorFlow as the backend. In this post, I go over the steps to install the same library with PyTorch as the backend. Many of the steps are similar but there are some important differences. I hope this…
-
Normalizing (Feature Scaling) Point Clouds for Machine Learning
Continuing my work on Machine Learning with point clouds in the realm of autonomous robots, and coming from working with image data, I was faced with the following question: does 3D data need normalization like image data does? The answer is a clear YES (duh!). Normalization, or feature scaling, is an important preprocessing step for many machine…
-
Simplifying Point Cloud Labeling with Contextual Images and Point Cloud Filtering
Annotating point clouds from multi-line 360° LiDAR is exceedingly difficult. Providing context in the form of camera frames and limiting the point cloud to the Field Of View (FOV) of the camera simplifies things. To achieve this, we first had to replace our old, and not so stable LiDAR mount, with a sturdier one capable…
-
Testing 3D Annotation Tools
Before you can train a supervised Deep Learning model, you must first label your data. Today I am testing Intel OpenVINO’s CVAT and MATLAB’s Lidar Labeler annotation tools for 3D data. First impressions, CVAT makes it easier to navigate the point cloud, but a small bug makes it hard to place the initial cuboid, making…
Soul Hackers Labs 