search

psychopa4 / PFNL

Progressive Fusion Video Super-Resolution Network via Exploiting Non-Local Spatio-Temporal Correlations
MIT License
138 stars 21 forks source link

how to implement 2x? #10

Open bonbert81 opened 4 years ago

psychopa4 commented 4 years ago

First, set self.scale=2. Second, make adjustment to forward(self, x). Only one tf.depth_to_space() is required for 2x. convmerge1=Conv2D(12, 3, strides=ds, padding='same', activation=activate, kernel_initializer=ki, name='convmerge1') merge=tf.concat(inp0,axis=-1) merge=convmerge1(merge) out=tf.depth_to_space(merge,2)

bonbert81 commented 4 years ago

thx for the answer. Should i change also self.eval_in_size to [256,480] ?