Open echeng-git opened 1 year ago
If we are using LiDAR data with images, then we can make modifications to the tool such that the user is no longer able to pan the camera (i.e. the center of rotation), rather, the center of rotation will move along the coordinates of each real world to provide a reference as the user annotates down the road.
Functionality for vehicle POV images should be removed, as I don't think that we are providing real images to go with our LiDAR data. Each bounding box, when manipulating a bounding box, is mapped to a real world image that has been provided (of the example dataset). We can remove these UI elements for now, or modify them if we are providing real-world images with our LiDAR data.
Relevant locations (of the code) of the image POV view (draggable, top of image) can be found from the following: In
index.html
:data/<YOUR_DATA_FOLDER>/camera/*/*.jpg
. Inpublic/js/editor.js
:imageContextManager
in this case, manages the POV images of the camera. Disabling this does remove the UI element, but also removes the feature of selecting the bounding boxes.public/js/image.js
.Relevant locations (of the code) of the image view (bottom) when selecting a bounding box can be found from the following: