English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

We present a pipeline of Image to Vector (Img2Vec) for masked image modeling(MIM) with deep features. To study which type of deep features is appropriatefor MIM as a learning target, we propose a simple MIM framework with serials ofwell-trained self-supervised models to convert an Image to a feature Vector asthe learning target of MIM, where the feature extractor is also known as ateacher model. Surprisingly, we empirically find that an MIM model benefitsmore from image features generated by some lighter models (e.g., ResNet-50,26M) than from those by a cumbersome teacher like Transformer-based models(e.g., ViT-Large, 307M). To analyze this remarkable phenomenon, we devise anovel attribute, token diversity, to evaluate the characteristics of generatedfeatures from different models. Token diversity measures the featuredissimilarity among different tokens. Through extensive experiments andvisualizations, we hypothesize that beyond the acknowledgment that a largemodel can improve MIM, a high token-diversity of a teacher model is alsocrucial. Based on the above discussion, Img2Vec adopts a teacher model withhigh token-diversity to generate image features. Img2Vec pre-trained onImageNet unlabeled data with ViT-B yields 85.1\% top-1 accuracy on fine-tuning.Moreover, we scale up Img2Vec on larger models, ViT-L and ViT-H, and get$86.7\%$ and $87.5\%$ accuracy respectively. It also achieves state-of-the-artresults on other downstream tasks, e.g., 51.8\% mAP on COCO and 50.7\% mIoU onADE20K. Img2Vec is a simple yet effective framework tailored to deep featureMIM learning, accomplishing superb comprehensive performance on representativevision tasks.

Img2Vec: A Teacher of High Token-Diversity Helps Masked AutoEncoders