Some code problems with frontend.py

experiencor / keras-yolo2

Easy training on custom dataset. Various backends (MobileNet and SqueezeNet) supported. A YOLO demo to detect raccoon run entirely in brower is accessible at https://git.io/vF7vI (not on Windows).

MIT License

1.73k stars 784 forks source link

Some code problems with frontend.py #247

Closed LBAWMY closed 6 years ago

LBAWMY commented 6 years ago

What does the line 190 do? class_mask = y_true[..., 4] tf.gather(self.class_wt, true_box_class) self.class_scale
Besides, I read the help doc about tf.gather(params,indices,), and it said the indices should be 0-D or1-D，and I test it in tensorflow, but I still didn't get it.

experiencor commented 6 years ago

it assigns different weights to different classes