search

HengLan / SMOT

[ECCV 2024] Beyond MOT: Semantic Multi-Object Tracking
Apache License 2.0
24 stars 0 forks source link

No module named 'smoter.data.datasets'. #2

Closed jxd31415 closed 2 months ago

jxd31415 commented 2 months ago

Thank you so much for providing a new direction for the MOT community! When trying to run your code, I encountered two problems that I would like the author to answer. 1. The datasets file is missing from smoter/data/. 2. Two images in the bensmot are invalid.

backyardknights commented 2 months ago

I previously encountered your problem 1, and I eliminated this error by moving a copy of the datasets folder from the built root directory to the corresponding folder. However, I don't think this solves the problem. Currently, running train_net.py can start the 0th iteration, but it shows that /smoter/data/smoter_dataset_dataloader.py cannot find the bensmot images, causing the program to interrupt and stop training. I'm not sure if the issue lies in your problem 1 or in smoter_dataset_dataloader.py itself.

backyardknights commented 2 months ago

I apologize for any inconvenience. I would like to know if you have resolved the issue and successfully replicated SMOT.

jxd31415 commented 2 months ago

I haven't solved the above problem and am still waiting for the author's reply.

Nathan-Li123 commented 2 months ago

Sorry for the late reply. Due to an error in the .gitignore file, there was an issue with missing files in the data folder. I have corrected this, and the corresponding files should now be available in the smoter/data folder. As for the second question, I don't actually understand what 'invalid images' mean, can you provide some specific error reports or detailed descriptions?

Nathan-Li123 commented 2 months ago

I previously encountered your problem 1, and I eliminated this error by moving a copy of the datasets folder from the built root directory to the corresponding folder. However, I don't think this solves the problem. Currently, running train_net.py can start the 0th iteration, but it shows that /smoter/data/smoter_dataset_dataloader.py cannot find the bensmot images, causing the program to interrupt and stop training. I'm not sure if the issue lies in your problem 1 or in smoter_dataset_dataloader.py itself.

I hope the changes to the repository will solve your problem, if not, please feel free to let me know.