FLIM Networks with Bag of Feature Points

2026-02-24

AI Summary

This work proposes FLIM-BoFP, a novel approach to address the high annotation dependency and training cost of convolutional neural networks in parasitic microscopy image detection. Departing from conventional layer-wise clustering strategies, FLIM-BoFP introduces a global Bag of Feature Points generated via one-time clustering at the input stage, which is then cross-layer mapped to each encoder layer to directly construct filters without backpropagation. Integrating an adaptive decoder with a clustering-driven encoder design, the method substantially enhances both the speed of filter estimation and the controllability of feature localization. Experimental results demonstrate that FLIM-BoFP outperforms FLIM-Cluster and current state-of-the-art methods in efficiency, accuracy, and generalization on parasitic optical microscopy image detection tasks.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Machine Learning: Multi-instance/Multi-view Learning
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text

Abstract

Convolutional networks require extensive image annotation, which can be costly and time-consuming. Feature Learning from Image Markers (FLIM) tackles this challenge by estimating encoder filters (i.e., kernel weights) from user-drawn markers on discriminative regions of a few representative images without traditional optimization. Such an encoder combined with an adaptive decoder comprises a FLIM network fully trained without backpropagation. Prior research has demonstrated their effectiveness in Salient Object Detection (SOD), being significantly lighter than existing lightweight models. This study revisits FLIM SOD and introduces FLIM-Bag of Feature Points (FLIM-BoFP), a considerably faster filter estimation method. The previous approach, FLIM-Cluster, derives filters through patch clustering at each encoder's block, leading to computational overhead and reduced control over filter locations. FLIM-BoFP streamlines this process by performing a single clustering at the input block, creating a bag of feature points, and defining filters directly from mapped feature points across all blocks. The paper evaluates the benefits in efficiency, effectiveness, and generalization of FLIM-BoFP compared to FLIM-Cluster and other state-of-the-art baselines for parasite detection in optical microscopy images.

Problem

• Salient Object Detection
• Few-shot Learning
• Filter Estimation
• Microscopy Image Analysis
• Annotation-efficient Learning

Innovation

• FLIM
• Bag of Feature Points
• filter estimation
• Salient Object Detection
• no backpropagation