gt-ripl / multiagentperception Goto Github PK

Sorry to bother you. Thanks for you work. When I enter the command 'python train.py --config configs/multi-request-multyi-support/mrms_when2com.yml --gpu=3' in the terminal to run the file, an error 'RuntimeError: CuDNN error: CUDNN_STATUS_MAPPING_ERROR' will be reported. Do you know how to solve this problem?

The onedrive link has again expired. Could you please update the link

Hello! Could you please tell me exactly the required version of the packages and python. In the tutorial, the required version of python is 3.6. In the requirements.txt, the required version of packages are
matplotlib==2.0.0
numpy==1.22.0
scipy==0.19.0
torch==0.4.1
torchvision==0.2.0
tqdm==4.11.2
pydensecrf
protobuf
tensorboardX
pyyaml
pretrainedmodels
opencv-python
But there are some version mismatch issues.
So I will appreciate it if you can tell me exactly the required version of the packages and python

“
File "test.py", line 14, in
from ptsemseg.visual import draw_bounding
ModuleNotFoundError: No module named 'ptsemseg.visual'
”

I can't import visual module！Could you please provide me with this module？ thank you very much

Sorry to bother you! Here I unziped "airsim-mrms-noise-data.zip", then I run the model, it throw out the error below:

Loaded: selection label.
Traceback (most recent call last):
  File "train.py", line 142, in <module>
    t_loader = data_loader(
  File "G:\TWang\MultiAgentPerception-master\ptsemseg\loader\airsim_loader.py", line 258, in __init__
    raise Exception(
Exception: No files for split=[train] found in dataset/airsim-mrms-noise-data

Actually, I split the dataset into the form, according to

Could you show me how you construct your dataset form? Or there lack the code of preprocess dataset to generate the split form?

Hi, can you show me how to access your AirSim-MAP datasets? I only got 404 not found error message back. Thanks a lot!

I find that the method of fusing the features is not the same as in the paper that relies on the attention mechanism score to select the features of the corresponding agent to fuse .
In the code, during training, the fused features are obtained directly using the attention mechanism, and it seems that the returned agent number, i.e., action_argmax , is not used subsequently.

https://github.com/GT-RIPL/MultiAgentPerception/blob/4ef300547a7f7af2676a034f7cf742b009f57d99/ptsemseg/trainer.py#L384C1-L398C1
And instead of using commun_label during training, it was used as a metric during validation and testing.
I feel that the log_action, action_argmax are really meaningless, and the features obtained during training are not all the features of a vehicle, but are obtained by using the attention method. Instead of selecting all the features of a certain car, in the test, new features are added after the attention is used.
https://github.com/GT-RIPL/MultiAgentPerception/blob/4ef300547a7f7af2676a034f7cf742b009f57d99/ptsemseg/models/agent.py#L629C1-L651C56

Hi,

The dataset link is expired, can you provide another link ?

Thanks in advance, keep up the hard work :)

The webpage indicates that the dataset is no longer available. Can you provide a new download link for the dataset?