GICN
GICN copied to clipboard
LiuShihHung
Missing key ['pc_indices'] in the dataset.
Thanks for your work. I notice that the key 'pc_indices' appears both in train and inference script, but missing in the h5 files.
also getting the same error. @yzheng97 were you able to figure out what the problem is?
KeyError: "Unable to open object (object 'pc_indices' doesn't exist)"
We haven't organized codes about generating dataset including key ['pc_indices'] yet. In actual, we just additionally record a new column pc_indices representing the index of this point in raw point cloud when generating cubes from raw point cloud.
@FishWantToFly So in the source code - pc_indices = fin['pc_indices'][block_id] # [4096] just represents an array from [0,...,4095]?
No, it actually represents the original point cloud index in raw point cloud, for sampled 4096 points (because we sample 4096 points per block)
Hi @FishWantToFly @LiuShihHung I work with @escanillans and we are trying to prove your method on some indoor scene data. I believe it would be best if we can generate validation resulting from your pre-trained model. Can you possible attache or link to a copy of the dataset with the pc_indices or the script used to generate these?
Same problem about the pc_indices, any help would be greatly appreciated!
pc_indicesisn't used in the code anyway. I just commented the line.
Then how do you run main_eval.py ?
Sorry, I was talking about train.py. I didn't run main_eval.py yet.
The h5 files are loaded successfully but I still have a "CUDA kernel failed" error message.
I don't know if pc_indices is really necessary. I defined it as: pc_indices = np.arange(4096) here, fixed some issues about squeezing numpy arrays, and main_eval.py finally worked.
Hi, @ammaradel .Could you please tell me how do you fix the issues about squeezing numpy arrays.I also meet the same error.
