search

chufengt / ALM-pedestrian-attribute

Code for the paper "Improving Pedestrian Attribute Recognition With Weakly-Supervised Multi-Scale Attribute-Specific Localization", ICCV 2019, http://arxiv.org/abs/1910.04562.
Apache License 2.0
188 stars 66 forks source link

Reproduce on RAP dataset #13

Closed jiajunzhang16 closed 4 years ago

jiajunzhang16 commented 4 years ago

hi, thanks for open source.

I tried to reproduce the results on RAP datasets.

There is quite a gap between paper and my reproduction.

Paper : Screenshot from 2020-05-25 14-07-16

My reproduction: Last epoch: Screenshot from 2020-05-25 14-15-12

The 10th epoch: Screenshot from 2020-05-25 14-22-40

I found the 10th epoch is much closer to the paper results.
How do you chose your best result ? (The best mA epoch or Recall epoch ?）

KaiiZhang commented 4 years ago

HI, i also want to reproduce on RAP dataset, I am troubled by a problem: I use the way in pedestrian-attribute-recognition-pytorch to process Rap dataset. I got two files: rap_partition.pkl and rap_partition.pkl These two files cannot be directly applied to this project because of different formats. How can I turn them into usable？ Can I have some suggestions? Thank you very much

chufengt commented 4 years ago

HI, i also want to reproduce on RAP dataset, I am troubled by a problem: I use the way in pedestrian-attribute-recognition-pytorch to process Rap dataset. I got two files: rap_partition.pkl and rap_partition.pkl These two files cannot be directly applied to this project because of different formats. How can I turn them into usable？ Can I have some suggestions? Thank you very much

refer to #4

chufengt commented 4 years ago

hi, thanks for open source.

I tried to reproduce the results on RAP datasets.

There is quite a gap between paper and my reproduction.

Paper : Screenshot from 2020-05-25 14-07-16

My reproduction: Last epoch: Screenshot from 2020-05-25 14-15-12

The 10th epoch: Screenshot from 2020-05-25 14-22-40

I found the 10th epoch is much closer to the paper results. How do you chose your best result ? (The best mA epoch or Recall epoch ?）

Did you correctly use the dataset and code? I achieve the results after about 30 epochs, refer to the Supplementary Material for more training detals.