In build_rpn.py, the rpn_conv2d_3x3 layer's kernel_size = [5, 5]? andstride = 1,so the output size will be smaller i think. is there any problems? maybe what i thought is wrong, i am not skillful to tf.
In the FPN paper, kernel size = 3, and the output size is the same as the input.
In
build_rpn.py, therpn_conv2d_3x3layer'skernel_size = [5, 5]? andstride = 1,so the output size will be smaller i think. is there any problems? maybe what i thought is wrong, i am not skillful to tf. In the FPN paper,kernel size = 3, and the output size is the same as the input.