search

WeijingShi / Point-GNN

Point-GNN: Graph Neural Network for 3D Object Detection in a Point Cloud, CVPR 2020.
MIT License
523 stars 114 forks source link

The output of kitti_native_evaluation #69

Open PuJanhan opened 3 years ago

PuJanhan commented 3 years ago

Thanks for your work, I modified model and trained it .I want to evaluate the model, But the output of kitti_native_evaluation is a empty file named plot . How do I evaluate the modified model?

WeijingShi commented 3 years ago

Hi @PuJanhan, Do you see any warnings or errors when you use kitti_native_evaluation? Does it give the evaluation results?

PuJanhan commented 3 years ago

Hi,I use python3 run.py checkpoints/car_auto_T3_train/ --dataset_root_dir DATASET_ROOT_DIR --output_dir DIR_TO_SAVE_RESULTS I get the results and Evaluate output results on the validation split, the result like this:

car_detection_AP : 0.699301, 0.699301, 0.699301 sh: 1:gnuplot: not found sh: 1:gnuplot: not found Error: /undefinedfilename in(car_detection_AP.eps) Operand stack:

Are the three values of car_detection_AP correspond to easy, moderate and hard in your paper? Why are they the same value? The validation split are all easy?

WeijingShi commented 3 years ago

Hi @PuJanhan,

  1. Can you check data subdirectory within DIR_TO_SAVE_RESULTS and see how many results files you get? you should see 3,769 files for the validation set. If not, check if the run.py finished without errors.
  2. You also need some dependencies to generate pdf. try to install them, re-compile kitti_native_evaluation.

Hope it helps. Thanks.

PuJanhan commented 3 years ago

Thanks, I follow your advice and success.

WeijingShi commented 3 years ago

Glad to know it works!

abhigoku10 commented 3 years ago

@PuJanhan from your modifications r u getting better results compared to the author ?? can you share the results obtained and the process applied