VDSR-Based
Approach:
Utilized a deep convolutional neural network for
super-resolution, making specific
architectural changes to reduce computational complexity while
maintaining
high-quality image
reconstruction.
GAN-Based
Approach:
Implemented a GAN framework to enhance the realism and
sharpness
of the
super-resolved images, with modifications aimed at reducing the
computational
load of the generator
and discriminator networks.
Key
Achievements:
Successfully reduced the computational complexity of both
VDSR
and GAN-based
models, making them more suitable for real-time or resource-constrained
environments.
Tools:
Python, TensorFlow/Keras, PyTorch.
