Embedded Files
In the past few decades, there have been significant advances in the field of deep learning and many different techniques have been introduced for Image to Image translation problems. Despite the presence of a number of techniques which perform well on different Image translation tasks, we do not have a comprehensive study comparing them or understanding the impact of introducing/changing any hyperparameter to the learning system.
We will define our evaluation metrics in a later section (Our Approach).
There are three aspects to the proposed project as outlined below.
Study of different aspects of the learning model for Image COlorization
Study of different aspects of the learning model for Image COlorization
We consider the following aspects of the learning system and report their relative performance when they are changed one at a time. We are training GANs while changing the following-
Performance COMPARISON of our best model against state of the art models
Performance COMPARISON of our best model against state of the art models
Our evaluation metric would be a closeness score (Defined in later section : Our Approach). We will be evaluating the score on the results of the test dataset for our best model as well as the existing state of the art models. We will present these results using their closeness score.
Training a model on an animation DAtaSET and run the model on a video
Training a model on an animation DAtaSET and run the model on a video
We plan to train a model using animation dataset and then run the corresponding model on an animation video. We will be doing this per frame basis. Therefore, this part would not have any quantitative measure of the result and there exists future possibility of applying other methods such as temporal consistency to enhance the results.
Page updated
Google Sites
Report abuse