# Just Noticeable Difference Modeling for Deep Visual Features

arXiv: https://arxiv.org/abs/2601.21933 · HTML: https://mlsanity.com/paper/2601.21933/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 26 of its 43 references that are in the map.
- First posted on arXiv 29 Jan 2026.

## Builds on (most cited first, 26 of 26)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://mlsanity.com/paper/2103.14030/) (2021)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [Mask R-CNN](https://mlsanity.com/paper/1703.06870/) (2017)
- [Towards Deep Learning Models Resistant to Adversarial Attacks](https://mlsanity.com/paper/1706.06083/) (2017)
- [Feature Pyramid Networks for Object Detection](https://mlsanity.com/paper/1612.03144/) (2016)
- [Fast R-CNN](https://mlsanity.com/paper/1504.08083/) (2015)
- [Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions](https://mlsanity.com/paper/2102.12122/) (2021)
- [Universal adversarial perturbations](https://mlsanity.com/paper/1610.08401/) (2016)
- [PVT v2: Improved Baselines with Pyramid Vision Transformer](https://mlsanity.com/paper/2106.13797/) (2021)
- [Black-box Adversarial Attacks with Limited Queries and Information](https://mlsanity.com/paper/1804.08598/) (2018)
- [Simple Black-box Adversarial Attacks](https://mlsanity.com/paper/1905.07121/) (2019)
- [BiLLM: Pushing the Limit of Post-Training Quantization for LLMs](https://mlsanity.com/paper/2402.04291/) (2024)
- [Model Breadcrumbs: Scaling Multi-Task Model Merging with Sparse Masks](https://mlsanity.com/paper/2312.06795/) (2023)
- [Video Coding for Machines: A Paradigm of Collaborative Compression and Intelligent Analytics](https://mlsanity.com/paper/2001.03569/) (2020)
- [Deep feature compression for collaborative object detection](https://mlsanity.com/paper/1802.03931/) (2018)
- [Synthesizing Robust Adversarial Examples](https://mlsanity.com/paper/1707.07397/) (2017)
- [USTC-TD: A Test Dataset and Benchmark for Image and Video Coding in 2020s](https://mlsanity.com/paper/2409.08481/) (2024)
- [Image Coding for Machines with Omnipotent Feature Learning](https://mlsanity.com/paper/2207.01932/) (2022)
- [LaWa: Using Latent Space for In-Generation Image Watermarking](https://mlsanity.com/paper/2408.05868/) (2024)
- [Just Noticeable Difference for Deep Machine Vision](https://mlsanity.com/paper/2102.08168/) (2021)
- [Pareto-Optimal Bit Allocation for Collaborative Intelligence](https://mlsanity.com/paper/2009.12430/) (2020)
- [Towards Top-Down Just Noticeable Difference Estimation of Natural Images](https://mlsanity.com/paper/2108.05058/) (2021)
- [End-to-End Learnable Multi-Scale Feature Compression for VCM](https://mlsanity.com/paper/2306.16670/) (2023)
- [Feature Coding in the Era of Large Models: Dataset, Test Conditions, and Benchmark](https://mlsanity.com/paper/2412.04307/) (2024)
- [DT-UFC: Universal Large Model Feature Coding via Peaky-to-Balanced Distribution Transformation](https://mlsanity.com/paper/2506.16495/) (2025)

## Built on by (most cited first, 3 of 3)

- [UC-VLM: Consistency-Driven Learning for AI-Generated Image Detection with Vision-Language Large Models](https://mlsanity.com/paper/2608.15238/) (2026)
- [Just Noticeable Difference Modeling for Token Compression in Vision-Language-Action Models](https://mlsanity.com/paper/2608.21247/) (2026)
- [See the Change, Keep the Flow: Unsupervised Action Segmentation via Spectral-Temporal Representation Learning](https://mlsanity.com/paper/2608.29611/) (2026)

## Abstract (arXiv)

Deep visual features are increasingly used as the interface in vision systems, motivating the need to describe feature characteristics and control feature quality for machine perception. Just noticeable difference (JND) characterizes the maximum imperceptible distortion for images under human or machine vision. Extending it to deep visual features naturally meets the above demand by providing a task-aligned tolerance boundary in feature space, offering a practical reference for controlling feature quality under constrained resources. We propose FeatJND, a task-aligned JND formulation that predicts the maximum tolerable per-feature perturbation map while preserving downstream task performance. We propose a FeatJND estimator at standardized split points and validate it across image classification, detection, and instance segmentation. Under matched distortion strength, FeatJND-based distortions consistently preserve higher task performance than unstructured Gaussian perturbations, and attribution visualizations suggest FeatJND can suppress non-critical feature regions. As an application, we further apply FeatJND to token-wise dynamic quantization and show that FeatJND-guided step-size allocation yields clear gains over random step-size permutation and global uniform step size under the same noise budget. Our code will be released after publication.