Research map / Detection and segmentation / CNN features
Reviving the Discarded High-Resolution Feature for Transformer-Based Tiny Object Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Revives the highest-resolution feature, discarded by DETR-like detectors for cost, to improve transformer-based tiny object detection while suppressing dense background noise.
Problem
Tiny object detection in DETR-like detectors suffers because fine details are lost in downsampling; the highest-resolution feature is usually discarded due to computational cost, and reintroducing it brings dense background noise.
Approach
Reintroduces the discarded highest-resolution feature into a transformer-based detector in a way that controls its computational cost and filters dense background noise, rather than naively adding it back.
Closest papers by meaning
- Learning 1-Bit Tiny Object Detector with Discriminative Feature Refinement (ICML 2024 · similarity 0.88)
- Sparse DETR: Efficient End-to-End Object Detection with Learnable Sparsity (ICLR 2022 · similarity 0.87)
- Trash to Treasure: Low-Light Object Detection via Decomposition-and-Aggregation (AAAI 2024 · similarity 0.84)
- Cubify Anything: Scaling Indoor 3D Object Detection (CVPR 2025 · similarity 0.84)
- Weakly Supervised Few-Shot Object Detection with DETR (AAAI 2024 · similarity 0.84)
- Not All Tokens Matter All The Time: Dynamic Token Aggregation Towards Efficient Detection Transformers (ICML 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.