search

chensong1995 / HybridPose

HybridPose: 6D Object Pose Estimation under Hybrid Representation (CVPR 2020)
MIT License
412 stars 64 forks source link

how to read into and make sense of data inside test_set_ape.npy #82

Closed monajalal closed 10 months ago

monajalal commented 10 months ago

After running this command: LD_LIBRARY_PATH=lib/regressor:$LD_LIBRARY_PATH python src/train_core.py

I get this file as output: /home/mona/HybridPose/output/linemod/test_set_ape.npy

and I am not sure how to read into this result to judge if using the pretrained weights is doing a good job or not. Is there an easier way to judge this like a single accuracy number?

monajalal commented 10 months ago

actually nvm I think something was wrong with using the pretrained weights saved file since evaluate is giving 0 for ADD-S score


(hybridpose) mona@mona-ThinkStation-P7:~/HybridPose$ python src/evaluate.py
ADD(-S) score of initial prediction is: 0.0
ADD(-S) score of final prediction is: 0.0
    # parser.add_argument('--prediction_file', type=str, default='output/occlusion_linemod/test_set_ape_40.npy')
    parser.add_argument('--prediction_file', type=str, default='/home/mona/HybridPose/output/linemod/test_set_ape.npy')
chensong1995 commented 10 months ago

Hi Mona,

Thanks for the question! My suggestion is to inspect the values of the record object here. The zero accuracy may be related to the other issue you raised #81 too.

I hope this helps! Let me know if you have further concerns.