Why is my data result so bad? Is it because of a problem with my operation, or is the quality of the point cloud not high enough？

[wiwiwisper]

Why is my data result so bad? Is it because of a problem with my operation, or is the quality of the point cloud not high enough？How was the demo dataset created

[ZhaoyangLyu]

Thank you for your interest in our research. There are several possible reasons for the bad result.

1. Check whether your point clouds bear the same scale and orientation as point clouds in our provided examples.
2. The model works better for point clouds sampled uniformly from meshes. It seems that your point clouds are not uniform.
3. The point clouds you provide are markedly deviant from that of our training data. In future research, we will endeavor to further enhance the generalization ability of MeshCoder.

[wiwiwisper]

Thank you for your response. I still have some doubts. I generated a simple box consisting of five rectangles and uniformly sampled the points, but I haven’t been able to achieve a good reconstruction result. The attachment contains the data I used. At 2025-11-18 10:08:01, "Zhaoyang Lyu" ***@***.***> wrote: ZhaoyangLyu left a comment (InternRobotics/MeshCoder#4) Thank you for your interest in our research. There are several possible reasons for the bad result. Check whether your point clouds bear the same scale and orientation as point clouds in our provided examples. The model works better for point clouds sampled uniformly from meshes. It seems that your point clouds are not uniform. The point clouds you provide are markedly deviant from that of our training data. In future research, we will endeavor to further enhance the generalization ability of MeshCoder. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[tathya-ow]

Do we require the point cloud in exact same orientation, like resting on the base from your dataset examples?