End-to-End Implicit Neural Representations for Classification

Implicit neural representations (INRs) such as NeRF and SIREN encode a signal in neural network parameters and show excellent results for signal reconstruction. Using INRs for downstream tasks, such as classification, is however not straightforward. Inherent symmetries in the parameters pose classification challenges and current works primarily focus on designing architectures that are equivariant to these symmetries. However, INR-based classification still significantly underperforms compared to pixel-based methods like CNNs. This work presents an end-to-end strategy for initializing SIRENs together with a learned learning-rate scheme to yield representations that improve classification accuracy. We show that a simple, straightforward, Transformer model applied to a meta-learned SIREN, without incorporating explicit symmetry equivariances, outperforms the current state-of-the-art. On the CIFAR-10 SIREN classification task, we improve the state-of-the-art from 38.8% to 60.1%. Moreover, we demonstrate scalability on the high-resolution Imagenette dataset achieving reasonable reconstruction quality with a classification accuracy of 60.94% while, to our knowledge, no other SIREN classification approach has managed to set a baseline for high-resolution images. We will make all code and results available.