Abstract
While parameter efficient tuning (PET) methods have shown great potential with transformer architecture on Natural Language Processing (NLP) tasks, their effectiveness is still under-studied with large-scale ConvNets on Computer Vision (CV) tasks. This paper proposes Conv-Adapter, a PET module designed for ConvNets. Conv-Adapter is light-weight, domain-transferable, and architecture-agnostic with generalized performance on different tasks. When transferring on downstream tasks, Conv-Adapter learns tasks-specific feature modulation to the intermediate representations of backbone while keeping the pre-trained parameters frozen. By introducing only a tiny amount of learnable parameters, e.g., only 3.5% full fine-tuning parameters of ResNet50, Conv-Adapter outperforms previous PET baseline methods and achieves comparable or surpasses the performance of full fine-tuning on 23 classification tasks of various domains. It also presents superior performance on few-shot classifications, with an average margin of 3.39%. Beyond classification, Conv-Adapter can generalize to detection and segmentation tasks with more than 50% reduction of parameters but comparable performance to the traditional full fine-tuning.
Abstract (translated)
URL
https://arxiv.org/abs/2208.07463
