Research map / Detection and segmentation / CNN features

MicroFiberDetect: An Application for the Detection of Microfibres in Wastewater Sludge Based on CNNs

AAAI 2025 · Detection and segmentation › CNN features · approach: Region-based and query-based detectors · 2 citations (OpenAlex)

Read the paperCheck your idea against 108,967 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.

Presents MicroFiberDetect, a desktop app that uses a fine-tuned Mask R-CNN to detect microfibres in sludge images and predict their colour and length.

Problem

Manual microscopy and chemical analysis for detecting and quantifying microfibres in wastewater sludge are labour-intensive, slow, and lack scalability.

Approach

Combines a Mask R-CNN instance segmenter with k-means colour clustering in HSL space and skeletonisation-based length measurement, packaged as a user-friendly desktop application for field experts without ML expertise.

Key results

• Mask R-CNN fine-tuned on two labelled sludge datasets (glass and cellulose acetate filters) with stratified five-fold cross-validation using AP as the main metric; detailed AP numbers are deferred to a separate journal submission.
• The desktop application processes an image in about 4 seconds on an Intel i7-8750H CPU.

Limitations

• Glass-filter images have organic matter interference, restricting detection to coloured fibres.
• Evaluation details and AP results are deferred to an external submission, so this paper gives few quantitative results.
• Length is estimated from skeleton pixel counts with a scale conversion, which the paper does not validate against ground truth.

Experimental setup

Datasets: Glass dataset (1,203 expert-labelled stereo-microscope images) CA dataset (1,287 expert-labelled stereo-microscope images) ImageNet (pretraining)

Metrics: average precision (AP) processing time per image

Baselines: U-Net + VGG16 microplastic pipeline (Lorenzo-Navarro et al. 2021)

Builds on: Mask R-CNN Detectron2 U-Net VGG16

Closest papers by meaning

• PHOEBI: An Open-World Benchmark for Bacterial Identification in Phase-Contrast Microscopy (NeurIPS 2026 · similarity 0.78)
• Automated State Estimation for Summarizing the Dynamics of Complex Urban Systems Using Representation Learning (AAAI 2024 · similarity 0.78)
• UrbanWaste: In-the-Bin Dataset for Waste Disposal Inspection with Multi-Granularity Hierarchical Labels (AAAI 2025 · similarity 0.77)
• MorphoHELM: A Comprehensive Benchmark for Evaluating Representations for Microscopy-Based Morphology Assays (NeurIPS 2026 · similarity 0.76)
• Infusing Synthetic Data with Real-World Patterns for Zero-Shot Material State Segmentation (NeurIPS 2024 · similarity 0.76)
• LAqua: Laplacian Pyramids for Aquatic Segmentation (Student Abstract) (AAAI 2025 · similarity 0.76)