Get amazing point clouds from your own pair of monocular cameras. The two cameras can even have different CMOS or different optics!

Deliverables (typically 1 week after job scheduled)

• Hammerhead. Point clouds using real-time algorithm.
• GroundTruth. Point clouds using offline algorithm.
• GridDetect. Occupancy maps using real-time algorithm.
• Presentation. Light analysis and QA, which includes point cloud snapshots, comparison to theoretical performance limits, and highlights of interesting features.

Input Data Requirements

For each stereo camera (a pair of cameras):

• Images (up to 10 scenarios with <100 images each)
• Synchronized image capture
• Camera extrinsic and intrinsic parameters
• The mounted tilt angle of the camera relative to the ground
• Distance to one object in the scene (preferably big and far) to do a one-time absolute range fine-tuning

The input format for the images, extrinsics, and intrinsics are defined here: https://cloud.nodarsensor.net/index.html#input-format

Step 1: Pay with a credit card

The quantity field refers to the number of stereo cameras that you would like tuned. If you would like us to overlay lidar, increase the quantity for each lidar that needs processing.

Step 2: Upload data

Upload your data to your server—Google Drive, Dropbox, S3, or other service—and provide us with access to the data.

Step 3: Email to notify