Alan

Results 14 comments of Alan

Questions on Training Configuration and Evaluation Errors for MTR Model on unScenes Dataset

Hi, your configuration looks good. nuscens has 3 splits: train, train_val, and val. We train MTR on train and validate on val. There are 32k samples in train and 9k...

Similar Wandb Bug Encountered (not able to be solved by commenting out wandb.log({"prediction": [wandb.Image(img)]})

I think these are just warnings which you don't need to worry about.

Clarification on Dataset Conversion & Trajectory Prediction in Unitraj

Can you try to comment the 'try, catch' in base_dataset.py. In this way you can see the errors in data loading.

Problem on NuScense Offical Evaluation

Hi, you need to convert the test data into ScenarioNet format. Please let me know if this doesn't work.

Questions about the implementation of trajectory prediction and training pipeline

Hi, the setting you mentioned is called 'scene-centric' prediction. You can find more information in this paper https://proceedings.neurips.cc/paper_files/paper/2023/file/b37c2e26b75ee02fcabd65a2a0367136-Paper-Conference.pdf

Questions about the implementation of trajectory prediction and training pipeline

Yes. We are working on implementing SMART with UniTraj.

Request for Preprocessed Dataset

Hi, Sorry for the late response. The processed dataset is much larger than the original, making it impossible for cluster-transferring. However I'm glad to answer your questions in data processing

Pip install fails due to deprecated nvidia-cuda-runtime-cu13

Still not working

TypeError: __init__() got an unexpected keyword argument 'verbose'

remove 'verbose in 'scheduler = MultiStepLR(optimizer, milestones=self.config['learning_rate_sched'], gamma=0.5,'

No module named 'waymo_open_dataset‘

try pip install waymo-open-dataset-tf-2-12-0==1.6.7