Vision] The Surprisingly Straightforward Scene Text...
2024-06-24
RoI Regression Loss : only consider text regions I_out(i) : output of the ith deconvolution pathway. (we use 3,4,5) M_i, I_gt(i) : box_mask, gt that was resize to I_out(i)*λ_i λ_i : 0.6, 0.8, 1.0 Perceptual...
Slow Learner