{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/genesis-v2-inferring-unordered-object","title":"GENESIS-V2: Inferring Unordered Object Representations without Iterative Refinement","arxiv_id":"2104.09958","date":"2021-04-20","proceeding":"NeurIPS 2021 12","authors":["Martin Engelcke","Oiwi Parker Jones","Ingmar Posner"],"abstract":"Advances in unsupervised learning of object-representations have culminated in the development of a broad range of methods for unsupervised object segmentation and interpretable object-centric scene generation. These methods, however, are limited to simulated and real-world datasets with limited visual complexity. Moreover, object representations are often inferred using RNNs which do not scale well to large images or iterative refinement which avoids imposing an unnatural ordering on objects in an image but requires the a priori initialisation of a fixed number of object representations. In contrast to established paradigms, this work proposes an embedding-based approach in which embeddings of pixels are clustered in a differentiable fashion using a stochastic stick-breaking process. Similar to iterative refinement, this clustering procedure also leads to randomly ordered object representations, but without the need of initialising a fixed number of clusters a priori. This is used to develop a new model, GENESIS-v2, which can infer a variable number of object representations without using RNNs or iterative refinement. We show that GENESIS-v2 performs strongly in comparison to recent baselines in terms of unsupervised image segmentation and object-centric scene generation on established synthetic datasets as well as more complex real-world datasets.","url_abs":"https://arxiv.org/abs/2104.09958v3","url_pdf":"https://arxiv.org/pdf/2104.09958v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"genesis-v2-inferring-unordered-object","repo_url":"https://github.com/applied-ai-lab/genesis","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"GPL-3.0"}},{"paper_slug":"genesis-v2-inferring-unordered-object","repo_url":"https://github.com/jinyangyuan/genesis","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"GPL-3.0"}}],"tasks":[{"task_slug":"clustering","task_name":"Clustering"},{"task_slug":"image-generation","task_name":"Image Generation"},{"task_slug":"image-segmentation","task_name":"Image Segmentation"},{"task_slug":"object","task_name":"Object"},{"task_slug":"representation-learning","task_name":"Representation Learning"},{"task_slug":"scene-generation","task_name":"Scene Generation"},{"task_slug":"unsupervised-image-segmentation","task_name":"Unsupervised Image Segmentation"},{"task_slug":"unsupervised-object-segmentation","task_name":"Unsupervised Object Segmentation"}],"methods":[{"method_slug":"adam","method_name":"Adam"},{"method_slug":"concatenated-skip-connection","method_name":"Concatenated Skip Connection"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"geco","method_name":"GECO"},{"method_slug":"group-normalization","method_name":"Group Normalization"},{"method_slug":"ic-sbp","method_name":"IC-SBP"},{"method_slug":"lstm","method_name":"LSTM"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"tanh-activation","method_name":"Tanh Activation"},{"method_slug":"transposed-convolution","method_name":"Transposed convolution"},{"method_slug":"u-net","method_name":"U-Net"}],"datasets_introduced":[],"methods_introduced":[{"slug":"ic-sbp","name":"IC-SBP","full_name":"Instance Colouring Stick-Breaking Process"}],"results":[{"leaderboard":"/sota/image-generation-on-objectsroom","task":"Image Generation","dataset":"ObjectsRoom","model":"GENESIS-V2","rank_in_archive_order":1,"of":3,"metrics":{"FID":"52.6"},"uses_additional_data":false},{"leaderboard":"/sota/image-generation-on-objectsroom","task":"Image Generation","dataset":"ObjectsRoom","model":"GENESIS","rank_in_archive_order":2,"of":3,"metrics":{"FID":"62.8"},"uses_additional_data":false},{"leaderboard":"/sota/image-generation-on-objectsroom","task":"Image Generation","dataset":"ObjectsRoom","model":"MONET-G","rank_in_archive_order":3,"of":3,"metrics":{"FID":"205.7"},"uses_additional_data":false},{"leaderboard":"/sota/image-generation-on-shapestacks","task":"Image Generation","dataset":"ShapeStacks","model":"GENESIS-V2","rank_in_archive_order":1,"of":3,"metrics":{"FID":"112.7"},"uses_additional_data":false},{"leaderboard":"/sota/image-generation-on-shapestacks","task":"Image Generation","dataset":"ShapeStacks","model":"GENESIS","rank_in_archive_order":2,"of":3,"metrics":{"FID":"186.8"},"uses_additional_data":false},{"leaderboard":"/sota/image-generation-on-shapestacks","task":"Image Generation","dataset":"ShapeStacks","model":"MONET-G","rank_in_archive_order":3,"of":3,"metrics":{"FID":"197.8"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on-1","task":"Unsupervised Object Segmentation","dataset":"ObjectsRoom","model":"GENESIS-V2","rank_in_archive_order":2,"of":5,"metrics":{"ARI-FG":"0.84"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on-1","task":"Unsupervised Object Segmentation","dataset":"ObjectsRoom","model":"SlotAttention","rank_in_archive_order":3,"of":5,"metrics":{"ARI-FG":"0.79"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on-1","task":"Unsupervised Object Segmentation","dataset":"ObjectsRoom","model":"GENESIS","rank_in_archive_order":4,"of":5,"metrics":{"ARI-FG":"0.63"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on-1","task":"Unsupervised Object Segmentation","dataset":"ObjectsRoom","model":"MONET-G","rank_in_archive_order":5,"of":5,"metrics":{"ARI-FG":"0.54"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on","task":"Unsupervised Object Segmentation","dataset":"ShapeStacks","model":"GENESIS-V2","rank_in_archive_order":2,"of":5,"metrics":{"ARI-FG":"0.81"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on","task":"Unsupervised Object Segmentation","dataset":"ShapeStacks","model":"SlotAttention","rank_in_archive_order":3,"of":5,"metrics":{"ARI-FG":"0.76"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on","task":"Unsupervised Object Segmentation","dataset":"ShapeStacks","model":"GENESIS","rank_in_archive_order":4,"of":5,"metrics":{"ARI-FG":"0.70"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on","task":"Unsupervised Object Segmentation","dataset":"ShapeStacks","model":"MONET-G","rank_in_archive_order":5,"of":5,"metrics":{"ARI-FG":"0.70"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on-shelf","task":"Unsupervised Object Segmentation","dataset":"Shelf&Tote Training Dataset","model":"GENESIS-V2","rank_in_archive_order":1,"of":4,"metrics":{"ARI":"0.55"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on-shelf","task":"Unsupervised Object Segmentation","dataset":"Shelf&Tote Training Dataset","model":"MONET-G","rank_in_archive_order":2,"of":4,"metrics":{"ARI":"0.11"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on-shelf","task":"Unsupervised Object Segmentation","dataset":"Shelf&Tote Training Dataset","model":"GENESIS","rank_in_archive_order":3,"of":4,"metrics":{"ARI":"0.04"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-object-segmentation-on-shelf","task":"Unsupervised Object Segmentation","dataset":"Shelf&Tote Training Dataset","model":"SlotAttention","rank_in_archive_order":4,"of":4,"metrics":{"ARI":"0.03"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2104.09958","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}