mlsanityData to 14 Oct 2026

Faster Segment Anything: Towards Lightweight SAM for Mobile Applications

arXiv 2306.14289 ↗First posted 25 Jun 2023cs.CVcited by 352

Not on the map. Why?

Abstract (arXiv)

Segment Anything Model (SAM) has attracted significant attention due to its impressive zero-shot transfer performance and high versatility for numerous vision applications (like image editing with fine-grained control). Many of such applications need to be run on resource-constraint edge devices, like mobile phones. In this work, we aim to make SAM mobile-friendly by replacing the heavyweight image encoder with a lightweight one. A naive way to train such a new SAM as in the original SAM paper leads to unsatisfactory performance, especially when limited training sources are available. We find that this is mainly caused by the coupled optimization of the image encoder and mask decoder, motivated by which we propose decoupled distillation. Concretely, we distill the knowledge from the heavy image encoder (ViT-H in the original SAM) to a lightweight image encoder, which can be automatically compatible with the mask decoder in the original SAM. The training can be completed on a single GPU within less than one day, and the resulting lightweight SAM is termed MobileSAM which is more than 60 times smaller yet performs on par with the original SAM. For inference speed, With a single GPU, MobileSAM runs around 10ms per image: 8ms on the image encoder and 4ms on the mask decoder. With superior performance, our MobileSAM is around 5 times faster than the concurrent FastSAM and 7 times smaller, making it more suitable for mobile applications. Moreover, we show that MobileSAM can run relatively smoothly on CPU. The code for our project is provided at hrefhttps://github.com/ChaoningZhang/MobileSAMtextcolorredMobileSAM), with a demo showing that MobileSAM can run relatively smoothly on CPU.

Key facts. Cited by 352 papers in the map (data through 14 Oct 2026). Builds on 35 of its 41 references that are in the map. First posted on arXiv 25 Jun 2023.

Builds on

35 of 41 references.

References · 35

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2020Language Models are Few-Shot Learnerscited by 32,419
  3. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  4. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  5. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  6. 2023Segment Anythingcited by 8,234
  7. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,513
  8. 2018MobileNetV2: Inverted Residuals and Linear Bottleneckscited by 4,795
  9. 2020Training data-efficient image transformers & distillation through attentioncited by 4,586
  10. 2021On the Opportunities and Risks of Foundation Modelscited by 4,033
  11. 2017MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applicationscited by 3,845
  12. 2023Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detectioncited by 3,332
23 more
  1. 2016V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentationcited by 2,292
  2. 2019Searching for MobileNetV3cited by 1,994
  3. 2017Focal Loss for Dense Object Detectioncited by 752
  4. 2021MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformercited by 642
  5. 2023Fast Segment Anythingcited by 290
  6. 2022EfficientFormer: Vision Transformers at MobileNet Speedcited by 241
  7. 2022TinyViT: Fast Pretraining Distillation for Small Vision Transformerscited by 234
  8. 2023Track Anything: Segment Anything Meets Videoscited by 234
  9. 2023Inpaint Anything: Segment Anything Meets Image Inpaintingcited by 233
  10. 2023EfficientViT: Memory Efficient Vision Transformer with Cascaded Group Attentioncited by 172
  11. 2023Can SAM Segment Anything? When SAM Meets Camouflaged Object Detectioncited by 100
  12. 2023Anything-3D: Towards Single-view Anything Reconstruction in the Wildcited by 92
  13. 2023A Complete Survey on Generative AI (AIGC): Is ChatGPT from GPT-4 to GPT-5 All You Need?cited by 84
  14. 2022How Does SimSiam Avoid Collapse Without Negative Samples? A Unified Understanding with Self-supervised Contrastive Learningcited by 65
  15. 2022Next-ViT: Next Generation Vision Transformer for Efficient Deployment in Realistic Industrial Scenarioscited by 65
  16. 2023Input Augmentation with SAM: Boosting Medical Image Segmentation with Segmentation Foundation Modelcited by 56
  17. 2023One Small Step for Generative AI, One Giant Leap for AGI: A Complete Survey on ChatGPT in AIGC Eracited by 55
  18. 2022Dual Temperature Helps Contrastive Learning Without Many Negative Samples: Towards Understanding and Simplifying MoCocited by 52
  19. 2023A Survey on Segment Anything Model (SAM): Vision Foundation Model Meets Prompt Engineeringcited by 45
  20. 2022Grad-StyleSpeech: Any-speaker Adaptive Text-to-Speech Synthesis with Diffusion Modelscited by 34
  21. 2023Segment Anything Model (SAM) Meets Glass: Mirror and Transparent Objects Cannot Be Easily Detectedcited by 34
  22. 2022Decoupled Adversarial Contrastive Learning for Self-supervised Adversarial Robustnesscited by 20
  23. 2023MP-FedCL: Multiprototype Federated Contrastive Learning for Edge Intelligencecited by 20

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 46; 2024: 126; 2025: 99; 2026: 81.

352 papers.

Sort:

By area:

  1. 2024Grounded SAM: Assembling Open-World Models for Diverse Visual Taskscited by 929
  2. 2023Gaussian Grouping: Segment and Edit Anything in 3D Scenescited by 393
  3. 2023VLFM: Vision-Language Frontier Maps for Zero-Shot Semantic Navigationcited by 205
  4. 2023Tracking Anything with Decoupled Video Segmentationcited by 197
  5. 2023EfficientSAM: Leveraged Masked Image Pretraining for Efficient Segment Anythingcited by 188
  6. 2023RepViT: Revisiting Mobile CNN From ViT Perspectivecited by 123
  7. 2023SAM-6D: Segment Anything Model Meets Zero-Shot 6D Object Pose Estimationcited by 117
  8. 2023SAM-CLIP: Merging Vision Foundation Models towards Semantic and Spatial Understandingcited by 111
  9. 2023Foundational Models Defining a New Era in Vision: A Survey and Outlookcited by 110
  10. 2023General In-Hand Object Rotation with Vision and Touchcited by 99
  11. 2023Segment Anything Meets Point Trackingcited by 81
  12. 2024Open-Vocabulary SAM: Segment and Recognize Twenty-thousand Classes Interactivelycited by 75
38 more
  1. 2024EVF-SAM: Early Vision-Language Fusion for Text-Prompted Segment Anything Modelcited by 65
  2. 2024Real-World Robot Applications of Foundation Models: A Reviewcited by 61
  3. 2024SAM2-Adapter: Evaluating & Adapting Segment Anything 2 in Downstream Tasks: Camouflage, Shadow, Medical Image Segmentation, and Morecited by 60
  4. 2023Adapting Segment Anything Model for Change Detection in HR Remote Sensing Imagescited by 56
  5. 2024Watermark Anything with Localized Messagescited by 53
  6. 2024LLM-Seg: Bridging Image Segmentation and Large Language Model Reasoningcited by 52
  7. 2025Perceive Anything: Recognize, Explain, Caption, and Segment Anything in Images and Videoscited by 51
  8. 2024VCP-CLIP: A visual context prompting model for zero-shot anomaly segmentationcited by 50
  9. 2025UAVs Meet LLMs: Overviews and Perspectives Toward Agentic Low-Altitude Mobilitycited by 49
  10. 2024FairMedFM: Fairness Benchmarking for Medical Imaging Foundation Modelscited by 47
  11. 2024DreamStory: Open-Domain Story Visualization by LLM-Guided Multi-Subject Consistent Diffusioncited by 47
  12. 2024IRSAM: Advancing Segment Anything Model for Infrared Small Target Detectioncited by 46
  13. 2023SAMPro3D: Locating SAM Prompts in 3D for Zero-Shot Instance Segmentationcited by 44
  14. 2023MobileSAMv2: Faster Segment Anything to Everythingcited by 42
  15. 2023EdgeSAM: Prompt-In-the-Loop Distillation for SAMcited by 41
  16. 2024Co-Speech Gesture Video Generation via Motion-Decoupled Diffusion Modelcited by 35
  17. 2024CorrCLIP: Reconstructing Patch Correlations in CLIP for Open-Vocabulary Semantic Segmentationcited by 34
  18. 2023TinySAM: Pushing the Envelope for Efficient Segment Anything Modelcited by 33
  19. 2024How to build the best medical image segmentation algorithm using foundation models: a comprehensive empirical study with Segment Anything Modelcited by 32
  20. 2024PTQ4SAM: Post-Training Quantization for Segment Anythingcited by 32
  21. 2024DarkSAM: Fooling Segment Anything Model to Segment Nothingcited by 32
  22. 2023Tokenize Anything via Promptingcited by 31
  23. 2024Touchstone Benchmark: Are We on the Right Way for Evaluating AI Algorithms for Medical Segmentation?cited by 31
  24. 2024SAGD: Boundary-Enhanced Segment Anything in 3D Gaussian via Gaussian Decompositioncited by 30
  25. 2024Hi-SAM: Marrying Segment Anything Model for Hierarchical Text Segmentationcited by 30
  26. 2024Moving Object Segmentation: All You Need Is SAM (and Flow)cited by 30
  27. 2023Neural feels with neural fields: Visuo-tactile perception for in-hand manipulationcited by 28
  28. 2025Fast-FoundationStereo: Real-Time Zero-Shot Stereo Matchingcited by 28
  29. 2024CAT-SAM: Conditional Tuning for Few-Shot Adaptation of Segment Anything Modelcited by 27
  30. 2023Cooperation Does Matter: Exploring Multi-Order Bilateral Relations for Audio-Visual Segmentationcited by 26
  31. 20243DitScene: Editing Any Scene via Language-guided Disentangled Gaussian Splattingcited by 25
  32. 2024PointSAM: Pointly-Supervised Segment Anything Model for Remote Sensing Imagescited by 24
  33. 2024Rethinking Interactive Image Segmentation with Low Latency, High Quality, and Diverse Promptscited by 22
  34. 2024g3D-LF: Generalizable 3D-Language Feature Fields for Embodied Taskscited by 22
  35. 2023EdgeFM: Leveraging Foundation Model for Open-set Learning on the Edgecited by 21
  36. 2024Subobject-level Image Tokenizationcited by 21
  37. 2024Efficient Track Anythingcited by 21
  38. 2025EdgeTAM: On-Device Track Anything Modelcited by 21

50 most cited of 352 All 352 (JSON)