autonomousvision
[Evaluation issue] Depth, Semantic image output issues and vehicle stopping
Thank you for providing the paper and code.
I have issues. Despite using pre-trained models, Ego vehicles do not move properly, often moving off the road or stalling.
So I have seen several posts discussing these issues and have activated visualization by adding the SAVE_PATH environment variable during the vehicle evaluation process.
- my gpu is nvidia titan rtx(24GB)
- (#135)
- (#128)
- ...
I confirmed the following problem.
HD Map, Depth, and Semantic segmentation images are output as follows.
I would like to inquire whether these issues are causing the vehicle not to move properly.
Looks like the models were not loaded properly. What path did you set for loading, and which files were in them?
Thank you so much for your response. The model path and its contents you mentioned are as follows: path : /home/tops/sschai/transfuser/model_ckpt/models_2022/transfuser/ and there are 4 files (args.txt, model_Seed1_39.pth, model_seed2_39.pth, model_seed_3_37,pth)
Additionally, I share the parts I modified and some of the paths.
hm I don't see an obvious issue. You need to debug here to check whether the model is loaded correctly.
As you pointed out, there was an issue with that variable. It seems that things got tangled up during code analysis. The problem has been resolved.
I appreciate your kind and helpful response and want to thank you once again from the bottom of my heart.