AI Paper Atlas

Research map / Detection and segmentation / CNN features

BiHashFormer: Hash-Driven Dual-Branch Transformer for Efficient Object Detection in HRW Shots

Jingchen Huang, Wenxi Li, Chenyang Lyu, Moran Liu, Haozhe Lin, Yuchen Guo

NeurIPS 2026 · Detection and segmentation › CNN features · approach: Efficient DETR query design · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes BiHashFormer, a hash-driven dual-branch transformer for efficient object detection in high-resolution wide (HRW) shots with sparse foreground.

Problem

Object detection in gigapixel-scale High-Resolution Wide (HRW) shots, where foreground is extremely sparse and interleaved with background, causing traditional detectors and attention to be inefficient and inaccurate.

Approach

A hash-driven dual-branch transformer that uses hashing to route/select foreground-relevant regions so attention avoids background, improving both efficiency and accuracy (details beyond the abstract not available).

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace SparseFormer backbone with BiHashFormerDINO head · PANDAAP50 / GFLOPs: AP50 0.780 -> 0.810; GFLOPs 75.71 -> 47.31Improved
replace Swin-T backbone with BiHashFormerDINO head · PANDAAP50 / GFLOPs: AP50 0.606 -> 0.810; GFLOPs 132.84 -> 47.31Improved
disable the compensation branch at inference (ablation)BiHashFormer · PANDAAP small: 0.591 -> 0.287Hurt
replace HashMiner with pooled classical attention for compensation (ablation)BiHashFormer · PANDAsmall-object efficiencyHurt
use random, MLP or gradient-based selection instead of HashMiner (ablation)compensation branch · PANDAAP smallHurt
swap order to K selects 50% Q instead of Q selects 50% K (ablation)HashAttention · PANDAAP50Hurt
use HashAttention instead of classical attention (Q to 100% K) (ablation)BiHashFormer · PANDAAP50 and costImproved
increase Hamming dimension k from 16 to 32 (ablation)BiHashFormer · PANDAAP50: 0.801 -> 0.810Improved
increase k from 32 to 64 (ablation)BiHashFormer · PANDAAP50: 0.810 -> 0.805No clear effect
use 3 prototype codes instead of 1 (ablation)HashMiner · PANDAAP small: 0.498 -> 0.591Improved
use 5 prototype codes instead of 3 (ablation)HashMiner · PANDAAP small: 0.591 -> 0.582No clear effect
use dynamic foreground-proportion RoI selection instead of fixed ratio (ablation)BiHashFormer vs SparseFormer · PANDAAP50Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy