SPTNet: An Efficient Alternative Framework for Generalized Category Discovery with Spatial Prompt Tuning

TL;DR

SPTNet introduces a novel two-stage adaptation framework with spatial prompt tuning for generalized category discovery, effectively improving classification accuracy by focusing on object parts and requiring minimal additional parameters.

Contribution

The paper proposes SPTNet, a new two-stage adaptation approach combining model fine-tuning and prompt learning, with a novel spatial prompt tuning method for better data representation alignment.

Findings

Achieves 61.4% accuracy on SSB benchmark, outperforming previous methods.

Uses only 0.117% of backbone parameters, demonstrating high efficiency.

Outperforms existing GCD methods by approximately 10% on average.

Abstract

Generalized Category Discovery (GCD) aims to classify unlabelled images from both `seen' and `unseen' classes by transferring knowledge from a set of labelled `seen' class images. A key theme in existing GCD approaches is adapting large-scale pre-trained models for the GCD task. An alternate perspective, however, is to adapt the data representation itself for better alignment with the pre-trained model. As such, in this paper, we introduce a two-stage adaptation approach termed SPTNet, which iteratively optimizes model parameters (i.e., model-finetuning) and data parameters (i.e., prompt learning). Furthermore, we propose a novel spatial prompt tuning method (SPT) which considers the spatial property of image data, enabling the method to better focus on object parts, which can transfer between seen and unseen classes. We thoroughly evaluate our SPTNet on standard benchmarks and demonstrate that our method outperforms existing GCD methods. Notably, we find our method achieves an average accuracy of 61.4% on the SSB, surpassing prior state-of-the-art methods by approximately 10%. The improvement is particularly remarkable as our method yields extra parameters amounting to only 0.117% of those in the backbone architecture. Project page: https://visual-ai.github.io/sptnet.