Image upsampling model. This model upasmple 1080p images into 2160p. The model was trained on only single image. However, it provides substantial performance in different images. Thanks to the generalizable property of convolution.
Model: "model" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= input (InputLayer) [(None, 1080, 1920, 3)] 0 conv2d_transpose_1 (Conv2DT (None, 2160, 3840, 8) 38408 ranspose) conv2d_transpose_2 (Conv2DT (None, 2160, 3840, 16) 51216 ranspose) conv2d_transpose_3 (Conv2DT (None, 2160, 3840, 16) 25616 ranspose) conv2d_transpose_4 (Conv2DT (None, 2160, 3840, 3) 1203 ranspose) tf.math.multiply_1 (TFOpLam (None, 2160, 3840, 3) 0 bda) tf.__operators__.add_1 (TFO (None, 2160, 3840, 3) 0 pLambda) ================================================================= Total params: 116,443 Trainable params: 116,443 Non-trainable params: 0 _________________________________________________________________
This Image is the only image that model was trained on: This image is completely unseen: