Issue with Road Extraction Using Deep Learning in Liadr360
haval abduljabbar sadeq
Hi,
I am facing an issue when trying to extract roads using the training model with deep learning in Liadr360. The processing for training model stops unexpectedly and doesnt continue, despite following the recommended steps. Here is what I’ve done:
- Converted the images to 8-bit format as required.
- Saved the annotations in a JSON file with the same name as the image.
- I have ensured that both the training and validation folders include the images and the corresponding annotation files.
However, the process still fails, and I get the error (as shown below). I have attached the training and validation folders, including the images and annotation files, for your review.
here is the link of the data https://drive.google.com/drive/folders/1M1-LxJhhOekKMwenuxUkarxqbl62acrV?usp=sharing
Could you please help me understand what might be causing this issue or if there’s something missing in my setup?
Thank you for your assistance!
[2025-01-14 18:37:47,259][lidar360Deeplearning.run][INFO] - Task Start
[2025-01-14 18:37:47,284][lidar360Deeplearning.utils.env_check][INFO] - GPU : NVIDIA GeForce RTX 2060
[2025-01-14 18:37:47,284][lidar360Deeplearning.utils.env_check][INFO] - Memory Total:6144 MB
[2025-01-14 18:37:47,284][lidar360Deeplearning.utils.env_check][INFO] - Memory Free: 4930.08984375 MB
[2025-01-14 18:37:47,284][lidar360Deeplearning.run][INFO] - Config Init
[2025-01-14 18:37:47,284][lidar360Deeplearning.trainer][INFO] - DEVICE : cuda
[2025-01-14 18:37:47,331][metrics][INFO] - Dataprepare Processing 1/1
[2025-01-14 18:38:11,451][metrics][INFO] - Processed file train_1, points number = 2820555
[2025-01-14 18:38:11,498][metrics][INFO] - Dataprepare Processing 1/1
[2025-01-14 18:38:15,373][metrics][INFO] - Processed file Validation_1, points number = 442873
[2025-01-14 18:38:15,388][lidar360Deeplearning.trainer][INFO] - Dataset generate success
[2025-01-14 18:38:15,514][lidar360Deeplearning.run][ERROR] - RUNNINGFAILURE,AttributeError Traceback (most recent call last):
File "C:\Program Files\GreenValley Suite\LiDAR360\8.1.0.0\python38\./Lib/site-packages\lidar360Deeplearning\run.py", line 94, in train
trainer = Trainer(config)
File "C:\Program Files\GreenValley Suite\LiDAR360\8.1.0.0\python38\./Lib/site-packages\lidar360Deeplearning\trainer.py", line 42, in __init__
self._initialize_trainer()
File "C:\Program Files\GreenValley Suite\LiDAR360\8.1.0.0\python38\./Lib/site-packages\lidar360Deeplearning\trainer.py", line 123, in _initialize_trainer
self._model: BaseModel = instantiate_model(copy.deepcopy(self._cfg), self._dataset)
File "C:\Program Files\GreenValley Suite\LiDAR360\8.1.0.0\python38\./Lib/site-packages\lidar360Deeplearning\models\model_factory.py", line 46, in instantiate_model
model = model_cls(model_config, "dummy", dataset, modellib)
File "C:\Program Files\GreenValley Suite\LiDAR360\8.1.0.0\python38\./Lib/site-packages\lidar360Deeplearning\models\segmentation\pvcnn.py", line 19, in __init__
super(PVCNN, self).__init__(option, dataset, sparse_class=True)
File "C:\Program Files\GreenValley Suite\LiDAR360\8.1.0.0\python38\./Lib/site-packages\lidar360Deeplearning\models\segmentation\base.py", line 77, in __init__
self.sampler = VoxelGrid(size=0.1) if not check_VoxelGrid(dataset) else None
File "C:\Program Files\GreenValley Suite\LiDAR360\8.1.0.0\python38\./Lib/site-packages\lidar360Deeplearning\utils\check_transform.py", line 8, in check_VoxelGrid
for transform in dataset.helper_transform.transforms:
AttributeError: 'Lidar360Dataset' object has no attribute 'helper_transform'
here is the link of the data https://drive.google.com/drive/folders/1M1-LxJhhOekKMwenuxUkarxqbl62acrV?usp=sharing
Solaire Guo
Dear Sir,
According to the log information, your current software version is 8.1.0.0. We recommend updating to version 8.1.1.0 and then attempting to train the model again. If the issue persists after the update, please feel free to contact us.
Best regards