Reproduce “LPT: Long-tailed Prompt Tuning for Image Classification” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Bowen Dong, Pan Zhou, Shuicheng Yan, Wangmeng Zuo
LPT adapts a frozen pretrained vision transformer for long-tailed image classification by learning shared prompts, then retrieving two additional prompts for each image using its features. On Places-LT, it reports 50.07% accuracy, compared with 42.82% for fully fine-tuned GCL using the same ImageNet-21k-pretrained ViT-B backbone. This comparison concerns accuracy and prompt parameters, not measured training time.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.