This is a showcase of my work in the course Automatic Image Analysis at the Technical University Berlin for my Master's in Computer Science 2023. I applied a few different neural network structures to the FashionMnist dataset for a denoising task.Specifically, I have designed a ...
- Base Model (a small Convolutional neural net in the classic sense)
- Batch Norm Model (The base model with batch norm in addition)
- Deeper Model (a deeper neural net with Batch Norm than the base model)
- Residual Model (a neural net with residual connections)
Original Images:
Noisy Images:
Reconstruced (Base Model):
Reconstructed (Batch Norm Model)
Reconstructed (Deeper Model)
Reconstructed (Residual Model)
Train Results:
