MatteoCappe / Matteo_thesis_workbook

0 stars 0 forks source link

OnePose basics #3

Open MatteoCappe opened 4 months ago

MatteoCappe commented 4 months ago

Paper - Code - Supplementary - Dataset

Pipeline and methods:

image

Inputs: Video scan sequence of the object and query images Output: Estimated object's pose

  1. Annotate the object's bounding box and the camera poses from the video scans in AR
  2. Reconstruct 3D sparse point cloud model by taking a video sequence of the object with Structure from Motion (SfM), while also constructing the 2D-3D correspondance graphs
  3. Train the GATs with them to directly find 2D-3D correspondance maps
  4. Test the model with a query image as input and find the 2D-3D correspondance map
  5. Use the map to solve the object pose estimation problem with Perspective-N-Point (PnP)
  6. Evaluate the results

Note: Check out the PowerPoint presentation to have more information on each step

Brief description of the techniques that are used:

Strenghts:

Limitations:

PowerPoint presentation:

OnePose PowerPoint presentation: OnePose.pptx See also Dope, as it uses synthetically-generated data to train its network: Dope.pptx Note: I will probably upload a presentation on OnePose++ and improve the one on OnePose in the future