TensorFlow implementation of Thin Plate Spline Spatial Transformer Network
# test.py
v = np.array([
[0.2, 0.2],
[0.4, 0.4],
[0.6, 0.6],
[0.8, 0.8]])
p = tf.constant(v.reshape([1, nx*ny, 2]), dtype=tf.float32)
t_img = tf.constant(img.reshape(shape), dtype=tf.float32)
t_img = TPS_STN(t_img, nx, ny, p, out_size)
- [Robust Scene Text Recognition with Automatic Rectification] (https://arxiv.org/abs/1603.03915)
- [Spatial Transformer Network] (https://arxiv.org/abs/1506.02025)
- TensorFlow STN implementation