FeatSharp: Your Vision Model Features, Sharper

TL;DR

FeatSharp introduces a cost-effective method to upsample low-resolution vision encoder features, enhancing detail preservation for improved performance in perception tasks and model training.

Contribution

It presents a novel technique for coherently upsampling vision encoder features, enabling better detail retention without high computational costs.

Findings

Improves feature map resolution in vision encoders.

Enhances performance on perception tasks.

Facilitates richer distillation targets.

Abstract

The feature maps of vision encoders are fundamental to myriad modern AI tasks, ranging from core perception algorithms (e.g. semantic segmentation, object detection, depth perception, etc.) to modern multimodal understanding in vision-language models (VLMs). Currently, in computer vision, the frontier of general purpose vision backbones is Vision Transformers (ViT), typically trained using contrastive loss (e.g. CLIP). A key problem with most off-the-shelf ViTs, particularly CLIP, is that these models are inflexibly low resolution. Most run at $224 \times 224$px, while the "high-resolution" versions are around $378-448$px, but still inflexible. We introduce a novel method to coherently and cheaply upsample the feature maps of low-resolution vision encoders while picking up on fine-grained details that would otherwise be lost due to resolution. We demonstrate the effectiveness of this approach on core perception tasks as well as within agglomerative model training using RADIO as a way of providing richer targets for distillation. Code available at https://github.com/NVlabs/FeatSharp .