List view
- No due date•2/3 issues closed
At the time of this milestone creation, the application supports addition of new nodes and ways to a TDEI workspace. It currently does not adhere to a specific schema, and instead focuses on addition information to aid research and mapping performance analysis. This milestone focuses on the following: 1. Support the OSW schema so that the data created from the application can be integrated with existing TDEI datasets. 2. Support update of existing nodes and ways (instead of just addition) so that the application can update existing TDEI datasets.
No due date- No due date•0/9 issues closed
Currently, the app hardcodes the workspace. Need to create a page that helps user selects the workspace they want to add to.
No due date•2/4 issues closedAnalyse segmented sidewalks to determine surface integrity. Need to look for potential cracks, breakages and gaps in sidewalks. Other Repositories: Python Ad-hoc: https://github.com/himanshunaidu/SurfaceIntegrityAnalysis iOS Ad-hoc: https://github.com/himanshunaidu/iOSSurfaceIntegrityChecker
No due date•5/14 issues closedImprove the CameraView to: - Make it orientation-flexible - Utilize AR-related features such as ARView - Capture the full camera and depth frames - Provision copies of the camera and depth frames and crops them (instead of cropping the source pixel buffers themselves and using them throughout), or provide cropped CIImage views of the frames - and more optimizations
No due date•6/18 issues closedImprove the point of interest location calculation algorithm. Use frameworks such as ARKit for metric scale calculations such as sidewalk width. ### Object Distance Calculation Take the image segment of the point of interest and get the depth of every pixel. Create the histogram of the depth of all the points involved, remove outliers, and return the mean depth. ### Object Location Calculation: 1. Replace the existing Flat-Earth approximation formula with the Great Circle formula While this makes an assumption that Earth is a perfect sphere, which is seemingly accurate enough to a reasonable extent. 2. Improve the object (relative) heading calculation using camera transform and intrinsics to take lateral position (field of view) of the objects into consideration. The current heading comes from Apple CoreLocation. Using the heading as is, makes the assumption that the object is perfectly straight in front of the device, which may not be the case. 3. Improve the object distance calculation using camera transform and intrinsics to take device tilt into consideration. The current distance value comes from the LiDAR depth map. Using the distance as is, makes the assumption that the device is perfectly upright, which is almost never the case. 4. Improve the object location using camera transform and intrinsics to take device rotation into consideration. ### Metric Scale Calculation 1. Use camera intrinsics and camera transform to calculate the sidewalk width
No due date•6/13 issues closed- Overdue by 4 month(s)•Due by June 30, 2025•8/12 issues closed
For non-Pro devices, we need to ensure that there is some kind of depth estimation process in place
No due date•2/4 issues closedMain Repository for this Milestone: https://github.com/himanshunaidu/ML_Pipeline_iOSPointMapper
No due date•8/12 issues closed- The geometry details of each POI need to be converted into a data object compatible with the OpenSideWalks schema. - The data object must be pushed to a common database (May use the TDEI API) Keywords: OpenSideWalks Schema, TDEI API
No due date•9/11 issues closedAfter the POI segments are: - identified by the app’s ContentView - validated by the user in the AnnotationView some specific POI types need post-processing to get their geometry (point and line details). E.g. Sidewalks: Break down into ways and nodes (Check OpenStreetMap data model)
No due date•31/39 issues closedOpen Issues: 1. We still need to define the user experience and interaction after the confirmation that "something's wrong here. I want to fix it". 2. Not clear how to pop back up to the AnnotationView once the interaction is done. Validation Types: - POI Class is correctly assigned - POI Class is incorrectly assigned - Assign the correct POI (probably create a Modal) - POI does not exist - (Optional) Multiple instances of the same class being detected as one instance - (Optional) POI has not been identified. Canvas feature to draw out the POI
No due date•5/15 issues closed