aboulch/ConvPoint

Nearest neighbor module error!

missbook520 opened this issue · 3 comments

hi ! Thanks for your work!
when I bulid the Nearest neighbor module and run test .py ,but i get the confusing result ,like [[[25769803779 ,5 ,0]]] ,and i set the K parameter equal to 3 ,i want to know why I get the so big data(trash data)
1
22

this module does not support windows I think

Hello sorry for late response,
As @Asher-1 said, the code has not been tested on Windows.
On Ubuntu, the following code seems to work:

batch_size = 1
num_points = 10
K = 3
pc = np.random.rand(batch_size, num_points, 3).astype(np.float32)
queries = np.ones((1,1,3), dtype=np.float32)
neigh_idx = nearest_neighbors.knn_batch(pc, queries, K, omp=True)
print("Queries", queries.shape)
print("PC", pc.shape)
print("Indices", neigh_idx.shape)

yeah,this work in ubuntu,thank you,Asher-1 and aboulch