{"url":"/method/snip","slug":"snip","name":"SNIP","full_name":"SNIP","full_name_withheld":false,"description_markdown":"**SNIP**, or **Scale Normalization for Image Pyramids**, is a multi-scale training scheme that selectively back-propagates the gradients of object instances of different sizes as a function of the image scale. SNIP is a modified version of MST where only the object instances that have a resolution close to the pre-training dataset, which is typically 224x224, are used for training the detector. In multi-scale training (MST), each image is observed at different resolutions therefore, at a high resolution (like 1400x2000) large objects are hard to classify and at a low resolution (like 480x800) small objects are hard to classify. Fortunately, each object instance appears at several different scales and some of those appearances fall in the desired scale range. In order to eliminate extreme scale objects, either too large or too small, training is only performed on objects that fall in the desired scale range and the remainder are simply ignored during back-propagation. Effectively, SNIP uses all the object instances during training, which helps capture all the variations in appearance and\r\npose, while reducing the domain-shift in the scale-space for the pre-trained network.","description_state":"present","introduced_year":null,"introduced_by":{"title":"An Analysis of Scale Invariance in Object Detection - SNIP","paper":"/paper/an-analysis-of-scale-invariance-in-object-1","first_author":"Bharat Singh","n_authors":2,"url_abs":null,"archive_paper_url":"https://paperswithcode.com/paper/an-analysis-of-scale-invariance-in-object-1"},"source":{"url":"http://arxiv.org/abs/1711.08189v2","title":"An Analysis of Scale Invariance in Object Detection - SNIP","url_on_a_paper_host":true},"code_snippet_url":null,"code_snippet_url_on_a_code_host":false,"categories":[{"area":"Computer Vision","area_id":"computer-vision","collection":"Multi-Scale Training","url":"/methods/category/multi-scale-training","pwc_aliases":[]}],"n_papers_tagged":17,"archive_num_papers":17,"papers_newest_first":[{"paper":null,"title":"DRIVE: Dual Gradient-Based Rapid Iterative Pruning","date":"2024-04-01","arxiv_id":"2404.03687","n_code_links":0,"syntology":null},{"paper":"/paper/grokking-tickets-lottery-tickets-accelerate","title":"Bridging Lottery Ticket and Grokking: Understanding Grokking from Inner Structure of Networks","date":"2023-10-30","arxiv_id":"2310.19470","n_code_links":1,"syntology":{"ran":3,"of":5,"unverified":2,"pointer_only":5}},{"paper":null,"title":"In defense of parameter sharing for model-compression","date":"2023-10-17","arxiv_id":"2310.11611","n_code_links":0,"syntology":null},{"paper":"/paper/snip-bridging-mathematical-symbolic-and","title":"SNIP: Bridging Mathematical Symbolic and Numeric Realms with Unified Pre-training","date":"2023-10-03","arxiv_id":"2310.02227","n_code_links":2,"syntology":{"ran":5,"of":5,"unverified":0,"pointer_only":1}},{"paper":null,"title":"Design of Discrete-time Matrix All-Pass Filters Using Subspace Nevanlinna Pick Interpolation","date":"2022-10-25","arxiv_id":"2210.14015","n_code_links":0,"syntology":null},{"paper":null,"title":"One-shot Network Pruning at Initialization with Discriminative Image Patches","date":"2022-09-13","arxiv_id":"2209.05683","n_code_links":0,"syntology":null},{"paper":"/paper/connectivity-matters-neural-network-pruning","title":"Connectivity Matters: Neural Network Pruning Through the Lens of Effective Sparsity","date":"2021-07-05","arxiv_id":"2107.02306","n_code_links":1,"syntology":{"ran":0,"of":12,"unverified":12,"pointer_only":0}},{"paper":null,"title":"Why is Pruning at Initialization Immune to Reinitializing and Shuffling?","date":"2021-07-05","arxiv_id":"2107.01808","n_code_links":0,"syntology":null},{"paper":null,"title":"How are journals cited? characterizing journal citations by type of citation","date":"2021-02-22","arxiv_id":"2102.11043","n_code_links":0,"syntology":null},{"paper":"/paper/scale-normalized-image-pyramids-with","title":"Scale Normalized Image Pyramids with AutoFocus for Object Detection","date":"2021-02-10","arxiv_id":"2102.05646","n_code_links":1,"syntology":null},{"paper":null,"title":"FreezeNet: Full Performance by Reduced Storage Costs","date":"2020-11-28","arxiv_id":"2011.14087","n_code_links":0,"syntology":null},{"paper":"/paper/pruning-redundant-mappings-in-transformer","title":"Pruning Redundant Mappings in Transformer Models via Spectral-Normalized Identity Prior","date":"2020-10-05","arxiv_id":"2010.01791","n_code_links":1,"syntology":null},{"paper":"/paper/winning-lottery-tickets-in-deep-generative","title":"Winning Lottery Tickets in Deep Generative Models","date":"2020-10-05","arxiv_id":"2010.02350","n_code_links":1,"syntology":null},{"paper":null,"title":"Pruning Neural Networks at Initialization: Why are We Missing the Mark?","date":"2020-09-18","arxiv_id":"2009.08576","n_code_links":0,"syntology":null},{"paper":null,"title":"Data-dependent Pruning to find the Winning Lottery Ticket","date":"2020-06-25","arxiv_id":"2006.14350","n_code_links":0,"syntology":null},{"paper":"/paper/pruning-via-iterative-ranking-of-sensitivity","title":"Pruning via Iterative Ranking of Sensitivity Statistics","date":"2020-06-01","arxiv_id":"2006.00896","n_code_links":1,"syntology":{"ran":1,"of":1,"unverified":0,"pointer_only":0}},{"paper":"/paper/an-analysis-of-scale-invariance-in-object-1","title":"An Analysis of Scale Invariance in Object Detection - SNIP","date":"2017-11-22","arxiv_id":"1711.08189","n_code_links":0,"syntology":null}],"papers_shown":17,"tasks":[{"task":"/task/network-pruning","name":"Network Pruning","papers":2},{"task":"/task/object","name":"Object","papers":2},{"task":"/task/object-detection","name":"Object Detection","papers":2},{"task":"/task/object-detection-1","name":"object-detection","papers":2},{"task":"/task/all","name":"All","papers":1},{"task":"/task/benchmarking","name":"Benchmarking","papers":1},{"task":"/task/contrastive-learning","name":"Contrastive Learning","papers":1},{"task":"/task/equation-discovery","name":"Equation Discovery","papers":1},{"task":"/task/few-shot-learning","name":"Few-Shot Learning","papers":1},{"task":"/task/image-classification","name":"Image Classification","papers":1},{"task":"/task/information-retrieval","name":"Information Retrieval","papers":1},{"task":"/task/math","name":"Math","papers":1},{"task":"/task/mathematical-induction","name":"Mathematical Induction","papers":1},{"task":"/task/mathematical-reasoning","name":"Mathematical Reasoning","papers":1},{"task":"/task/memorization","name":"Memorization","papers":1},{"task":"/task/model-compression","name":"Model Compression","papers":1},{"task":"/task/property-prediction","name":"Property Prediction","papers":1},{"task":"/task/retrieval","name":"Retrieval","papers":1},{"task":"/task/sensitivity","name":"Sensitivity","papers":1},{"task":"/task/symbolic-regression","name":"Symbolic Regression","papers":1}],"tasks_shown":20,"n_tasks":21,"usage_by_year":[{"year":"2017","papers":1},{"year":"2020","papers":6},{"year":"2021","papers":4},{"year":"2022","papers":2},{"year":"2023","papers":3},{"year":"2024","papers":1}],"row_source":"methods_table","archive":{"source":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","archive_url":"https://paperswithcode.com/method/snip"},"syntology_read_at":"2026-09-24T18:15:14+00:00"}