ProtoQuant: Quantization of Prototypical Parts For General and Fine-Grained Image Classification
This addresses the challenge of interpretability and generalization in image classification for researchers and practitioners, though it is incremental as it builds on existing prototypical-parts methods.
The paper tackled the problem of prototype drift and computational expense in prototypical parts-based models for image classification by introducing ProtoQuant, which uses latent vector quantization to stabilize prototypes and achieve competitive accuracy on ImageNet and fine-grained benchmarks like CUB-200 and Cars-196.
Prototypical parts-based models offer a "this looks like that" paradigm for intrinsic interpretability, yet they typically struggle with ImageNet-scale generalization and often require computationally expensive backbone finetuning. Furthermore, existing methods frequently suffer from "prototype drift," where learned prototypes lack tangible grounding in the training distribution and change their activation under small perturbations. We present ProtoQuant, a novel architecture that achieves prototype stability and grounded interpretability through latent vector quantization. By constraining prototypes to a discrete learned codebook within the latent space, we ensure they remain faithful representations of the training data without the need to update the backbone. This design allows ProtoQuant to function as an efficient, interpretable head that scales to large-scale datasets. We evaluate ProtoQuant on ImageNet and several fine-grained benchmarks (CUB-200, Cars-196). Our results demonstrate that ProtoQuant achieves competitive classification accuracy while generalizing to ImageNet and comparable interpretability metrics to other prototypical-parts-based methods.