aicps/roadscene2vec

A Tool for Extracting and Embedding Road Scene-Graphs

roadscene2vec's Issues

use_case_1 bug

When I run use_case_1.py, the following error appears, what is the problem?
/home/lcx/anaconda3/envs/av3/bin/python /home/lcx/PycharmProjects/roadscene2vec-main/examples/use_case_1.py
0it [00:00, ?it/s]
0it [00:00, ?it/s]

• finished

about new pretrained models

Hi, I have downloaded your newly updated folder 'use_case_data', but I didn't find the 'sequence_classification_example_model.pt'.
I would like to know whether the folder 'use_case_data' of the previous version can still be used.

Use_case_2.py error

roadscene2vec/learning/util/trainer.py", line 40, in init
if self.config.training_configuration["seed"] != None:
AttributeError: 'configuration' object has no attribute 'training_configuration'

pytorch: 1.10.0
torchvision: 0.11.1
cudatoolkit: 11.3.1

Some questions about raw data

Hello, thank you for providing excellent work about scene-graph! I am interested in your work, and I‘d like to ask that, could you provide the raw data in google drive? The source about IEEE dataport can not be downloaded free. It is more helpful to promote works in this field. Thank you again！

I reported an error when trying to run the first use case.

I reported an error when trying to run the first use case. Does this problem exist in a specific version of networkx?
pytorch==1.10.0
cudatoolkit==11.3.1
networkx==2.6.3

The following is error infomation:
(av) ranyabing@ranyabing:~/roadscene2vec/examples$ python use_case_1.py
100%|██████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:02<00:00, 2.45s/it]
0%| | 0/1 [00:00<?, ?it/s]
Traceback (most recent call last):
File "/home/ranyabing/roadscene2vec/examples/use_case_1.py", line 35, in
visualize(scenegraph_extraction_config) #visualize extracted scenegraphs
File "/home/ranyabing/roadscene2vec/roadscene2vec/util/visualizer.py", line 172, in visualize
visualize_real_image(extraction_config)
File "/home/ranyabing/roadscene2vec/roadscene2vec/util/visualizer.py", line 192, in visualize_real_image
draw(extractor, sequence[frame], bbox, bev, sg, save_path='output.png')
File "/home/ranyabing/roadscene2vec/roadscene2vec/util/visualizer.py", line 131, in draw
sg_img = draw_scenegraph_pydot(sg)
File "/home/ranyabing/roadscene2vec/roadscene2vec/util/visualizer.py", line 105, in draw_scenegraph_pydot
A = nx_pydot.to_pydot(sg)
File "/home/ranyabing/anaconda3/envs/av/lib/python3.9/site-packages/networkx/drawing/nx_pydot.py", line 263, in to_pydot
raise ValueError(
ValueError: Node names and attributes should not contain ":" unless they are quoted with "". For example the string 'attribute:data1' should be written as '"attribute:data1"'. Please refer pydot/pydot#258

Annotation Tool

Dear editor,
Thank you for your brilliant work. According to section 3.1 of your paper, your implement an annotation tool with a graphical user interface (GUI). However, I can not find the corresponding code in this project. Could you provide more information that how to use the annotation tool？
Looking forward to your reply.

pretrained model

Hi！Thank you for your excellent work. In terms of use case 2, how can I get the pretrained model? I didn't find '/pretrained_models/mrgcn_sequence_classification_model.pt'.

pretrained model was wrong

Hi, I found the pretrained model that you provide is not right. I use Netron to visualize the pretrained model, model is incomplete， the FastRGCNConv module is lost. And I found the def save_model(self) in trainer.py is wrong, the function can not save the whole model weight, could you please fix it, thank you.

Graphviz installation requirement not explicitly mentioned

Hello,

I tried running the first use_case after performing all the installation steps as mentioned in your README. I faced the following error:

model_final_f10217.pkl: 178MB [00:18, 9.71MB/s]                                                                                                                                                                                                                                           
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:03<00:00,  3.04s/it]
  0%|                                                                                                                                                                                                                                                               | 0/1 [00:01<?, ?it/s]
Traceback (most recent call last):
  File "/home/bourne/anaconda3/envs/av/lib/python3.9/site-packages/pydot.py", line 1923, in create
    stdout_data, stderr_data, process = call_graphviz(
  File "/home/bourne/anaconda3/envs/av/lib/python3.9/site-packages/pydot.py", line 132, in call_graphviz
    process = subprocess.Popen(
  File "/home/bourne/anaconda3/envs/av/lib/python3.9/subprocess.py", line 951, in __init__
    self._execute_child(args, executable, preexec_fn, close_fds,
  File "/home/bourne/anaconda3/envs/av/lib/python3.9/subprocess.py", line 1821, in _execute_child
    raise child_exception_type(errno_num, err_msg, err_filename)
FileNotFoundError: [Errno 2] No such file or directory: 'dot'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/bourne/dlr/git_repos/roadscene2vec/examples/use_case_1.py", line 33, in <module>
    visualize(scenegraph_extraction_config) #visualize extracted scenegraphs
  File "/home/bourne/dlr/git_repos/roadscene2vec/roadscene2vec/util/visualizer.py", line 168, in visualize
    visualize_real_image(extraction_config)
  File "/home/bourne/dlr/git_repos/roadscene2vec/roadscene2vec/util/visualizer.py", line 188, in visualize_real_image
    draw(extractor, sequence[frame], bbox, bev, sg, save_path='output.png')
  File "/home/bourne/dlr/git_repos/roadscene2vec/roadscene2vec/util/visualizer.py", line 127, in draw
    sg_img = draw_scenegraph_pydot(sg)
  File "/home/bourne/dlr/git_repos/roadscene2vec/roadscene2vec/util/visualizer.py", line 103, in draw_scenegraph_pydot
    img = A.create_png()
  File "/home/bourne/anaconda3/envs/av/lib/python3.9/site-packages/pydot.py", line 1733, in new_method
    return self.create(
  File "/home/bourne/anaconda3/envs/av/lib/python3.9/site-packages/pydot.py", line 1933, in create
    raise OSError(*args)
FileNotFoundError: [Errno 2] "dot" not found in path.

This was resolved by installing graphviz conda install -c anaconda graphviz. However I failed to find any place in the documentation where this was explicitly mentioned as a requirement, which it should be.

where is the bev.json file come from?

I can not find it in the HDD dataset