You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Input 2D Graph to a completely different (nodes, edges, coordinates) output 1D graph but with morphological similarity.
Graphs can be single-loop-graph for now, multiloop ones later. Mask or negative weight on hidden edge.
Need to come up with a way to generate geometric-graph embedding that will convolute at nodes (having coordinates) around arcs (having geometry, say, a set of point coordinates), then pool them to form a meaningful representation.
3DPoseLite: Stick to Mesh, pose estimation, 2 encoder 1 decoder VAE method, need to reverse to make it dimensions reduction.
Input 2D Graph to a completely different (nodes, edges, coordinates) output 1D graph but with morphological similarity.
Graphs can be single-loop-graph for now, multiloop ones later. Mask or negative weight on hidden edge.
Need to come up with a way to generate geometric-graph embedding that will convolute at nodes (having coordinates) around arcs (having geometry, say, a set of point coordinates), then pool them to form a meaningful representation.
3DPoseLite: Stick to Mesh, pose estimation, 2 encoder 1 decoder VAE method, need to reverse to make it dimensions reduction.
https://openaccess.thecvf.com/content/WACV2021/papers/Dani_3DPoseLite_A_Compact_3D_Pose_Estimation_Using_Node_Embeddings_WACV_2021_paper.pdf
Real crux would be how to formulate pooling to aggregate all incident curves info plus node coordinates info into a single number!!
The text was updated successfully, but these errors were encountered: