IMC-Net: A Lightweight Content-Conditioned Encoder with Multi-Pass Processing for Image Classification

TL;DR

IMC-Net introduces a lightweight, content-conditioned multi-pass encoder for image classification that balances accuracy and efficiency by adaptively processing feature map regions, achieving competitive results with fewer resources.

Contribution

The paper proposes a novel multi-pass, content-conditioned encoder architecture that reduces computational cost while maintaining high accuracy without heavy auxiliary modules or pretraining.

Findings

Achieves competitive accuracy with fewer parameters and FLOPs.

Faster inference compared to similar-sized models.

Effective transferability across multiple datasets.

Abstract

We present a compact encoder for image categorization that emphasizes computation economy through content-conditioned multi-pass processing. The model employs a single lightweight core block that can be re-applied a small number of times, while a simple score-based selector decides whether further passes are beneficial for each region unit in the feature map. This design provides input-conditioned depth without introducing heavy auxiliary modules or specialized pretraining. On standard benchmarks, the approach attains competitive accuracy with reduced parameters, lower floating-point operations, and faster inference compared to similarly sized baselines. The method keeps the architecture minimal, implements module reuse to control footprint, and preserves stable training via mild regularization on selection scores. We discuss implementation choices for efficient masking, pass control, and representation caching, and show that the multi-pass strategy transfers well to several datasets without requiring task-specific customization.