relationships.txt and classes_nuscenes.txt
I see you've been updating during the National Day holiday - you've really been working hard! I'm extremely interested in this project, and your work is truly impressive. I've read the paper over a dozen times and am currently going through the codebase while attempting to reproduce it.
However, I'm currently stuck because I'm missing these two files, particularly relationships.txt. My understanding is that relationships.txt defines the spatial correlations between objects (left/right/front/back etc.), is that correct?
The file paths in question are: relationship_file_path='../lidargen/dataset/transforms_3d/scene_graph/relationships.txt', classnames_file_path='../lidargen/dataset/transforms_3d/scene_graph/classes_nuscenes.txt'
I've created my own version of classes_nuscenes.txt. like this, is it correct?
car truck construction_vehicle bus trailer barrier motorcycle bicycle pedestrian traffic_cone
I'm very eager to get this project running locally as soon as possible. Would it be possible for you to release the complete code soon, or at least provide these two files so I can continue my work?
Thank you for your consideration!
Oh, So sorry for the issue. Yes, you are right. The details of these files are in following: ./lidargen/dataset/transforms_3d/scene_graph ├── classes_nuscenes.txt ├── init.py ├── pycache │ ├── init.cpython-310.pyc │ └── scene_graph.cpython-310.pyc ├── relationships.txt └── scene_graph.py
relationships.txt: left right front behind close by bigger than smaller than taller than shorter than
classes_nuscenes.txt: ego car truck construction_vehicle bus trailer motorcycle bicycle pedestrian
Thank for your interesting again.
Thank you for your response. The issues with relationships.txt and classes_nuscenes.txt have been resolved. However, I am now missing the file
pretrained_models/nuscenes-object-1000000.pth,
and I couldn’t find it on Hugging Face either. Could you please help?
The same issue: missing the file of pretrained_models/nuscenes-object-1000000.pth,
Sorry for that, file can be downloaded from: huggingface. But it's not necessery in current version of codebase, each condition can extract directly from open-sourced dataset (NuScenes for this work.)