Why are all the results zero?

[jgyy4775]

The command that I use "python main.py --config-file configs/sgg_res101_joint.yaml --inference --resume 0". And I use this checkpoints: https://drive.google.com/file/d/1VbgX3_OLhzgdNseXgS_2DiLmJ8qiG8P/view

What am I doing wrong?...

[jungjun9150]

I think you set the wrong checkpoint path. What is the checkpoint file path and file name?

[jgyy4775]

@jungjun9150 File path is /media/ailab/c414d3e4-ba21-486b-8c45-0c1e072e32fa/home/l/GY/graph-rcnn.pytorch/checkpoints/vg_benchmark_object/R-101-C4/faster_rcnn/BatchSize_6/Base_LR_0.005/sg_baseline_ckpt.pth

File name is sg_baseline_ckpt.pth

[jnhwkim]

The argument of resume should be the number of iterations you trained and the file name should be changed accordingly in my understanding. Please refer to https://github.com/jwyang/graph-rcnn.pytorch/blob/master/lib/scene_parser/rcnn/utils/checkpoint.py#L54 .

[jgyy4775]

@jnhwkim @jungjun9150 what mean "0:scratch, 1:resume or 2:finetune" and "inference"? I dont know "scratch" , "resume" and "infernece" means.

[jnhwkim]

@jgyy4775 Sorry but could you elaborate your question?

[jgyy4775]

@jnhwkim
In main.py

"scratch" , "resume" , "inference" is command line option. what's that mean? train이나 test시 입력하는 resume, inference의 의미가 궁금합니다. 어떤역할을 하는 것들 인가요??

[jungjun9150]

--inference: It is to take a test your model. When you go to the main.py, you can see the line that if --inference do test() else train(). no argument. just store true or false.

--resume : you have to set the filename to be equal --resume's argument.

[jgyy4775]

@jungjun9150 So what is scratch??

[jnhwkim]

I think args.mode is not used and deprecated. Could you confirm this, @jwyang ?