How and Why to train Image to Image deep learning models

Image credits : shutterbug

Introduction

But WHY ?

And HOW ?

Image credits : www.educba.com
  1. Autoencoder based CNN Model
  2. Pyramid Real Image Denoising Network (PridNET)

Input pipeline

Model Architecture : Auto encoder based CNN

Model Architecture : PRIDNet

  • First, the noise estimation stage uses channel attention mechanism to recalibrate the channel importance of input noise.
  • Second, at the multi-scale denoising stage, pyramid pooling is utilized to extract multi-scale features.
  • Third, the stage of feature fusion adopts a kernel selecting operation to adaptively fuse multi-scale features.
Image credits : https://arxiv.org/abs/1908.00273?context=cs.CV

Model Training

Inference

Output after 10 epochs — watermark has not gone completely but it has blended into the image and its intensity is quite low
Output after 10 epochs — watermark has not gone completely but it has blended into the image and its intensity is quite low

Final Notes

Link to Code

References

--

--

--

ML Engineer, AI enthusiast. darshil3011@gmail.com www.thinkinbytes.in

Love podcasts or audiobooks? Learn on the go with our new app.

Recommended from Medium

Support Vector Machines — the basics

What is gradient descent(intuitively)?

On the shortcomings of continuous representations of chemical space

Comparing RNN and CNN models on invoice extraction: LSTM vs GRU vs TCN

Brand Logo Detection in Images : GCS + Cloud Run (Docker image) + gcloud (Part-3)

How To Create An Opensource NLU API With Rasa

Journey through the World of NLP -NLP Pipeline! — Part-2.1

DeFlow: Learning Complex Image Degradations from Unpaired Data with Conditional Flows

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Darshil

Darshil

ML Engineer, AI enthusiast. darshil3011@gmail.com www.thinkinbytes.in

More from Medium

Visualizing Neural Network Activation

Triplet Loss on ImageNet Dataset

Observing behavior of tokens in Visual Transformers

My Tensorflow collection