visualcomputinginstitute / 3d-semantic-segmentation Goto Github PK

Thanks for sharing your code. Could you also give a pretrained model or instructions for the outdoor(vKITTI) dataset as well? It would help me a lot.

I am not getting what should be the data path ex if it is gru model and the configuration file is area_1 then which area's path should be places in data path and what should be the corresponding test_set, also can we specify more than 1 test_set, if yes how? How is the data split for train and validation and how to test the model on unlabelled data set?

Hi, I'm very interested in your project, I wonder if your project supports inputting a 3D point cloud and outputting a semantically segmented 3D point cloud (i.e., with each point cloud labeled)

Thank you for your excellent code, but I have a problem with the import of the Dataset.
Error that I encounter when I follow your Readme.
I downloaded the Dataset and I use the file path located on my machine, but without success for you to tell me what element I have to modify?
ValueError : Dataset could not be found under' '

Thank you very much for your great work ssp+spg, can you tell me the miou of each category of your work on the vkitti dataset? If you can tell me, I will cite your paper and compare it with your work in our experiment. Looking forward to your reply.

Thanks for providing the code for public. I am trying to create npy file for training. I am getting the following error. can you please help me to fix it?

(3DSeg) C:\Users\Documents\3DDeep_LIDAR\tools>python prepare_s3dis.py --input_dir Dataset --output_dir OUT
Program was launched with the following arguments:
~ input_dir Dataset
~ output_dir OUT

0%| | 0/272 [00:00<?, ?it/s]
Traceback (most recent call last):
File "prepare_s3dis.py", line 75, in
main(params)
File "prepare_s3dis.py", line 57, in main
out_filename = elements[-3] + '_' + elements[-2] + '.npy' # Area_1_hallway_1.npy
IndexError: list index out of range

Running evaluation epoch 0001 / 0061: 0it [00:00, ?it/s]
Traceback (most recent call last):
File "run.py", line 329, in
main(config, log_dir, isTrain)
File "run.py", line 48, in main
eval_one_epoch(sess, ops, model, dataset, epoch, config['train']['epochs'])
File "run.py", line 237, in eval_one_epoch
avg_loss = loss_sum / float(total_seen / model.batch_generator.num_points)
ZeroDivisionError: float division by zero

While training the standford_indoor data at running training epoch 001/061, I got this error. How could i fix this error?

It has been two years since you published the code. When can you publish the code on the vkitti data set?We're looking forward to it.

Hello,
thanks for the great work,
how did you train on vkitti3d data?
thanks

I want to perform classification and segmentation of outdoor pointclouds using the vkitty dataset and your 3d-semantic-segmentation tool.

After downloading the dataset and prepare the dataset using your tools, I want to train a model.

Which model should I use to do this ? gru_neighbor, multi_block or multi_scale_cu ?

Thanks for sharing your work !

Thanks for providing the code for public，you have done the experiment with only XYZ input features, without RGB in your paper，can you tell me how to complete the experiment without RGB? because the dataset which i download is constitute of XYZ and RGB.

Hello and thanks for this great work !

I am following your procedure, I downloaded the Standford dataset and I am wondering a few things:

• Do we have to manually create your directory structure with the files from the datasets or do you have a way to automate it ?
• Same for .npy files since the dataset offers .txt files ?

Thanks