T-Rex

model paper
IDEA Lab
Interactive object counting and detection via visual prompting. Users mark points or boxes on reference images, then T-Rex detects all similar objects. T-Rex2 synergizes text and visual prompts via contrastive learning. State-of-the-art on class-agnostic counting benchmarks.
Paper (T-Rex, arXiv)Paper (T-Rex2, arXiv)GitHubProject Page

Outputs 3

T-Rex

model

Interactive object counting model using visual prompts for detecting and counting any objects with zero-shot capabilities.

GitHubProject Page

T-Rex: Counting by Visual Prompting

paper

Formulates object counting as open-set detection with visual prompts and interactive refinement.

Paper (arXiv)

arXiv: 2311.13596

T-Rex2: Towards Generic Object Detection via Text-Visual Prompt Synergy

paper

Extends T-Rex by synergizing text and visual prompts within a single model through contrastive learning for generic object detection.

Paper (arXiv)

arXiv: 2403.14610

Venue: ECCV 2024

visioncountingopen-source