Retention-Constrained Post-Training Quantization of Cellpose-SAM for Stem Cell Microscopy
Summary
The paper evaluates compression schemes for Cellpose-SAM in stem cell microscopy, demonstrating that mixed-precision quantization achieves 6.76x reduction without catastrophic failures while maintaining segmentation accuracy.
View Cached Full Text
Cached at: 09/21/26, 11:20 AM
Paper page - Retention-Constrained Post-Training Quantization of Cellpose-SAM for Stem Cell Microscopy
Source: https://huggingface.co/papers/2609.21038
https://huggingface.co/login?next=%2Fpapers%2F2609.21038- 
Abstract
Inducedpluripotentstemcell(iPSC)cultureincreasinglyreliesonsegmentationfoundationmodels,yetdeploymentonlaboratoryCPUsandedgehardwarerequirescompressionschemesthatarebothefficientandauditable.Wepresentadeployment-orientedevaluationofcompressedCellpose-SAMusingapre-specifiedretentioncriterion:the95%cluster-bootstrapintervalofmeanchangefromFP32mustremainaboveafixed-0.02marginforeveryimagingmodality.Onastratified176-fieldpanelspanningBBBC038nuclei,BBBC039U2OSfluorescence,andNISTiPSCimagesacrossdensityregimes,weight-onlyW8A16preservesinstanceF1acrossallmodalities.Asensitivity-guidedmixedW4/W8scheme,usingfourINT8exceptions,achievesa6.76xreductioninweightstoragewithnoobservedcatastrophicfailures(0/176fields),matchingW8A16atthissamplesize.Incontrast,ternaryweight-onlyquantizationachieves12.08xcompressionbutfailscatastrophicallyon169/176fields.Theseresultsdemonstratethatcompressionshouldbeevaluatedbymodality-stratifieddownstreamretentionratherthansingle-numberaccuracy,andestablishareproducibleprotocolforauditingcompressedfoundationmodelsinregulatedstem-cellimaging.
View arXiv pageView PDFAdd to collection
Community
Paper author
Paper submitter
Cellpose-SAM is used for segmenting cells and nuclei in microscopy images. Deploying it directly on laboratory instruments would enable analysis at the point of acquisition, but these devices have limited compute and memory compared with specialized workstations.
We therefore asked whether Cellpose-SAM could be compressed for local deployment without compromising segmentation accuracy, and whether compression affected imaging conditions differently. Cellpose-SAM compressed 12-fold degraded on 169 of 176 image fields, while a conservative mixed-precision scheme achieved a 6.76× reduction with no observed failures. These results show that compression must be evaluated across imaging conditions, not by model size or average accuracy alone.
Upload images, audio, and videos by dragging in the text input, pasting, orclicking here.
Tap or paste here to upload images
https://huggingface.co/login?next=%2Fpapers%2F2609.21038-
Get this paper in your agent:
hf papers read 2609\.21038
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper5
#### capicu-ai/cellpose-sam-wquant-w4a16-g64 Image Segmentation• 0.2B• Updatedabout 10 hours ago • 17
#### capicu-ai/cellpose-sam-wquant-w8a8-qdq Image Segmentation• 0.3B• Updatedabout 10 hours ago • 25
#### capicu-ai/cellpose-sam-wquant-w8a16 Image Segmentation• 0.3B• Updatedabout 10 hours ago • 19
#### capicu-ai/cellpose-sam-wquant-mixed-w4-w8 Image Segmentation• 0.2B• Updatedabout 10 hours ago • 19
Browse 5 models citing this paper## Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2609.21038 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2609.21038 in a Space README.md to link it from this page.
Collections including this paper1
Similar Articles
@MaximeRivest: this is a really cool and powerful trick for image segmentation with sam without prompting! > You can exploit the small…
A technique is shared for using the Segment Anything Model (SAM) to segment images without prompting by exploiting similarities in microscopy images, enabling zero-shot mask propagation similar to video.
@lillyguisnet: In 2023 I used a combination of SAM1 + a custom vit classifier to ID and automate the segmentation of my microscopy ima…
The author shares a method to automate microscopy image segmentation using SAM1 and SAM3.1, achieving 100% accuracy by grouping similar images and exploiting video-like propagation.
E-PMQ: Expert-Guided Post-Merge Quantization with Merged-Weight Anchoring
This paper introduces E-PMQ, an expert-guided post-merge quantization framework that addresses the combined deviations from merging and quantization, achieving significant accuracy improvements on multi-task merged models like CLIP-ViT and FLAN-T5.
Joint Structural Pruning and Mixed-Precision Quantization for LLM Compression
A novel end-to-end framework for LLM compression that jointly optimizes structural pruning and mixed-precision quantization, achieving significant perplexity reductions and speedups over state-of-the-art methods, especially at ultra-low bit precisions.
ReQuant: Fixed-Grid Discrete Refinement for Post-Training Quantization
ReQuant introduces a backpropagation-free, fixed-grid discrete refinement stage for post-training quantization (PTQ) that iteratively improves initial quantized models while preserving the quantized format, showing consistent gains across various LLMs and bit-widths.