Checkpoint 3 Reflection (overwritten when writing the final):
By checkpoint 3, we had written most of the code for the project. We had experimented with both a GAN and an autoencoder-based model and had found that the autencoder model worked significantly better than the GAN model. We had improved the effectiveness of our model by adding residual layers between the encoder and decoder of our model. We were not absolutely certain why this improved our model, but we had added them thinking that they would pass more information about details in our image to the decoder, resulting in a more robust reconstruction.
We had the outline for the poster created, and had populated it with an introduction, information about our dataset/problem, and some of the methods section. We had some intermediate results, but did not include them in the poster at that time.
We had delegated the work relatively evenly up to that point, and continued delegating the work evenly for the rest of the duration of the project.
Log in or sign up for Devpost to join the conversation.