mcmingchang


[1.zip](https://github.com/isl-org/Open3D/files/9098034/1.zip) This is my data

This is my solution def inverse_matrix(m): m_np = np.array(m) m_np_inv = np.linalg.inv(m_np) return m_np_inv import trimesh mat, whd = trimesh.bounds.oriented_bounds(np.array(inlier_cloud.points)) mat_inv = inverse_matrix(mat) bbox = trimesh.creation.box(whd, mat_inv) new_obb = o3d.utility.Vector3dVector(np.array(bbox.vertices))...

  The panoramic segmentation effect is good, but the instance segmentation effect is obviously poor. I don't know if the object is too small

model: channels: 32 num_blocks: 7 semantic_classes: 6 instance_classes: 6 sem2ins_classes: [] semantic_only: False # T为预训练 ignore_label: -100 grouping_cfg: score_thr: 0.2 radius: 0.04 # 用于分组的K近邻的搜索半径。 越精细越小 与scale有关 mean_active: 500 # 用于约束K-NN后的总大小...

  semantic_labels and instance_labels

the black points are invalid，should I add a forecast category?

How can I optimize the recognition the small facets in instance segmentation

x4_split will affect the prediction accuracy?

But I have some big data，can it be converted to CPU calculation?