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

About evaluation on validation split #67

Open arxidinakbar opened 3 years ago

arxidinakbar commented 3 years ago

Hi, thank you for the open source , such an inspirational work. And there is some question about the reproducing of the codes. When I running evaluation_object_offline, occur error message as below:

evaluate_object_offline DATASET_ROOT_DIR/labels/training/label_2/ /home/arxidin/Point-GNN/output/
evaluate_object_offline: command not found

Is it wrong typing of my command line, or I miss some important steps.

Best luck

arxidinakbar commented 3 years ago

@WeijingShi Updates: found the mistake in command line(rookie mistake), the right call should be: ./evaluate_object_offline DATASET_ROOT_DIR/labels/training/label_2/ /home/arxidin/Point-GNN/output/ And new error occurred when running the evaluation code, as below:

ERROR: Couldn't read: 000231.txt of ground truth. Please write me an email!
An error occured while processing your results.

I checked out the 000231.txt, seems its ok to me:

Car -1 -1 0 132.33486895156815 181.5794514345332 248.27443998017648 233.97209698790462 1.560591 1.7343118 4.3006825 -14.027986 1.8830698 24.448862 1.4830902 102.07324069049855 
Car -1 -1 0 504.69382647689747 179.5086588049625 547.112850544539 213.6784762864795 1.4785603 1.607873 3.6240041 -3.859847 1.8066678 33.75335 1.5581328 43.10357801273189 
Car -1 -1 0 559.0860858880757 181.67230286550253 612.0103256153467 227.90150760268764 1.3648738 1.6127851 3.408605 -0.8023536 1.6757717 23.701895 1.5288224 44.05179057416208 
Car -1 -1 0 310.49684756678374 182.580857395939 354.8517716061144 210.40833217113317 1.4670331 1.5533721 4.149431 -15.944693 2.058747 41.72419 1.4171709 7.9335449299411875 
Car -1 -1 0 832.1130615306412 182.52353926509488 875.4255568075206 201.5944788865052 1.5814065 1.5984839 4.1575165 21.571075 2.4603262 63.67595 -0.6624789 2.672199348913683

Wish get some tips for this issue.

Best luck

arxidinakbar commented 3 years ago

Hi, found another mistake in command line that result folder dir were imcomplete, sorry for the careless operation. The _evaluation_objectoffline ran successfully, but the plot folder contain no output, may you drop some hints on this issue?

Best luck

WeijingShi commented 3 years ago

Hi @arxidinakbar, Glad to know you can run evaluation_object_offline successfully. Did you get the evaluation results? if so, during the progress, did you see any warnings regard to the pdf generation? The tool may need some extra dependence to generate pdf.

arxidinakbar commented 3 years ago

Hi @WeijingShi, Its nice to get your reply. I didn't pay much attention on warning information where I got the evaluation results successfully. And found no log file of evaluation. Wondering that which kind of tools may needed for PDF generation, thought I should install pre-packages properly.( I would check my virtual-envs to make sure I've installed all packages that you refered)

Best luck

PuJanhan commented 3 years ago

Hi , I place the file DIR_TO_SAVE_RESULTS under kitti_native_evaluation, when I use ./evaluate_object_offline DATASET_ROOT_DIR/labels/training/label_2/ DIR_TO_SAVE_RESULTS/data I meet the same problem : the file plot is empt How do you solve the problem?

Skye2099 commented 2 years ago

Hi , I place the file DIR_TO_SAVE_RESULTS under kitti_native_evaluation, when I use ./evaluate_object_offline DATASET_ROOT_DIR/labels/training/label_2/ DIR_TO_SAVE_RESULTS/data I meet the same problem : the file plot is empt How do you solve the problem?

hi, the output dir should not include 'data', just DIR_TO_SAVE_RESULTS

phoebedddd commented 1 year ago

Hi @arxidinakbar, Glad to know you can run evaluation_object_offline successfully. Did you get the evaluation results? if so, during the progress, did you see any warnings regard to the pdf generation? The tool may need some extra dependence to generate pdf.

Hi @WeijingShi,I see the warnings regard to the pdf generation,can you tell me what's the extra dependence?and How can generate estimates of pedestrians and bicycles?Why did I only get an evaluation of the car?I am looking forward to your reply. Thank you

phoebedddd commented 1 year ago

Hi @arxidinakbar, Glad to know you can run evaluation_object_offline successfully. Did you get the evaluation results? if so, during the progress, did you see any warnings regard to the pdf generation? The tool may need some extra dependence to generate pdf.

Hi @WeijingShi,I see the warnings regard to the pdf generation,can you tell me what's the extra dependence?and How can generate estimates of pedestrians and bicycles?Why did I only get an evaluation of the car?I am looking forward to your reply. Thank you

I already know the answer to the above question from reading your previous answers, but I still have a question. What should I change if I want to train pedestrians and bicycles separately?

jiubie89 commented 8 months ago

Hi,

Hi, found another mistake in command line that result folder dir were imcomplete, sorry for the careless operation. The _evaluation_objectoffline ran successfully, but the plot folder contain no output, may you drop some hints on this issue?

Best luck

Can you please specify the folder directory you changed at that time? I encountered the same issue and tried changing the folder directory, but it didn't work. I want to refer to what you did and compare it with my attempts to modify.

Best luck