Point-GNN.pytorch
Point-GNN.pytorch copied to clipboard
It takes me about 12mins to train the Car model for each epoch with GTX 1070 (8G) using the TF version(https://github.com/WeijingShi/Point-GNN), but it takes me about 20mins to train the...
@Shudeng @Wu-Shudeng in the training pipeline of mytrain.py in line batch = data_provider.provide_batch([1545, 1546]) y is it specified 1545,1546 can you clarify on this , if i am training the...
 Can I just use Pytorch?
@Shudeng hi thanks for sharing the code base i have few queries 1. can you we run the inference on the cpu 2. can you please share the trained model...
Hi, thank you for your implementation of pytorch version for the paper. Strangely, I meet the following problem when running train.sh ***** Error in `python3': free(): invalid pointer: 0x000055e59b686ba8 ***...
How can we evaluate this model and visualize results (same function as run.py) ? Thanks
What is the difference between encoded_boxes and valid_boxes obtained in a batch?
Thank you very much for your work, it is very useful for my study! Do I only need to run python3 mytrain.py configs/car_auto_T3_train_train_config configs/car_auto_T3_train_config --dataset_root_dir data for training?
I think there is a mistake in the model.py: 117, where `point_set_keypoint_coordinates = point_features[point_set_keypoint_indices[:, 0]]` should be: `point_set_keypoint_coordinates = point_coordinates[point_set_keypoint_indices[:, 0]]` @Shudeng