Mini toolsLingBot Vision Small
By Robbyant
LingBot-Vision-Small is the smallest variant (ViT-S/16) in the LingBot-Vision family of self-supervised Vision Transformer backbones for dense spatial perception. It is pretrained via masked boundary modeling, a boundary-centric objective where teacher-discovered boundary tokens are forced into the masked set and receive both semantic self-distillation and categorical boundary-field supervision, producing patch features that preserve object boundaries, shapes, and semantic regions. The checkpoint contains backbone weights only, released as a PyTorch model.pt file. Intended uses include dense feature visualization via PCA, image feature extraction, and backbone initialization for downstream dense prediction research such as depth estimation, semantic segmentation, video object segmentation, and depth completion.
New
Image
Overview
LingBot-Vision-Small is a lightweight Vision Transformer (ViT-S/16) backbone pretrained with self-supervised masked boundary modeling for dense spatial perception. It produces boundary-aware patch token embeddings for feature extraction, PCA visualization, and downstream dense prediction tasks such as depth estimation and semantic segmentation.
Last updated: July 8, 2026
