Hugues THOMAS

icon indicating a website link https://huguesthomas.github.io icon indicating an email [email protected]

icon company University of Toronto icon location Toronto icon location AI/robotics Researcher. Postdoc at University of Toronto. Focus: Deep Learning and 3D Point clouds. Indoor navigation

Results 160 comments of Hugues THOMAS

test_model

Both strategies would work indeed

Out of memory when using test_models.py

See previous issues #24

Issues about dataset

> print(len(training_dataset[0])) what is `training_dataset` in this case? If it is the one defined in the training script, you should not use training_dataset[i] to access data this is not the...

visualize_deformations "list index out of range"

for the first bug, it seems the code does not see any deformable convolution in the network. For the second bug, I cannot help much as it is not a...

config for different datasets

The config for different datasets is mostly the same, only some data related parameters should change. There are numbers of issues where I already talked about config parameters. You can...

config for different datasets

The point features are in the outputs of the network: https://github.com/HuguesTHOMAS/KPConv-PyTorch/blob/73e444d486cd6cb56122c3dd410e51c734064cfe/utils/trainer.py#L188 You can see them used in the validation function: https://github.com/HuguesTHOMAS/KPConv-PyTorch/blob/73e444d486cd6cb56122c3dd410e51c734064cfe/utils/trainer.py#L480-L484

ValueError: Truth values are stored in a 0D array instead of 1D array

Seems you are providing an empty `truth` array in the confusion matrix function. You can try to track what happens by following the code backwards. The truth arrays come from...

Question on testing part segmentation on custom data

Hi @tengyu-liu, In `tester.py`, follow the execution of the code in the `test_cloud_segmentation` function. On the dataset side, what you want to follow are the functions `get_batch_gen` and `get_tf_mapping`. The...

Preparing own data for regression with multiple outputs

Hi @nikogamulin, Thank you for your interest in my work. This problem of yours seems very interesting and hard to tackle. First I think that the success or failure of...

How to get the inside values of the python closure function `def spatially_regular_gen()` and `def tf_map()`?

Hi @Violetit, Sorry for the late reply, `spatially_regular_gen()` function is a python generator so you should be able to use a simple python print to see the values in there....