Megvii-BaseDetection / DisAlign

Implementation of "Distribution Alignment: A Unified Framework for Long-tail Visual Recognition"(CVPR 2021)
Apache License 2.0
117 stars 10 forks source link

ABOUT file "num_shots_v0.5.npy". #5

Closed Liu-wanbing closed 3 years ago

Liu-wanbing commented 3 years ago

Dear Author: Hi, is the file "num_shots_v0.5.npy" necessary for training? Can you provide the file?https://github.com/MegviiBaseDetection/DisAlign/blob/eb85ca08d2204c612e5ef1c54534d49a688dc838/segmentation/lvis0.5/mask_rcnn/res50/mask_rcnn.res50.fpn.lvis.multiscale.cos_norm.disalign.1x/roi_heads/fast_rcnn.py#L45

def _init_weights(self, exp_scale=0.7):
    import os
    import cvpods
    import numpy as np
    folder_name = "/".join(cvpods.__path__[0].split('/')[:-1])
    num_samples_list = np.load(os.path.join(folder_name, "datasets/lvis/num_shots_v0.5.npy"))
tonysy commented 3 years ago

I have updated the readme and uploaded the related files. Feel free to re-open if you have any questions.