{"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/ensembling-off-the-shelf-models-for-gan","title":"Ensembling Off-the-shelf Models for GAN Training","arxiv_id":"2112.09130","date":"2021-12-16","proceeding":"CVPR 2022 1","authors":["Nupur Kumari","Richard Zhang","Eli Shechtman","Jun-Yan Zhu"],"abstract":"The advent of large-scale training has produced a cornucopia of powerful visual recognition models. However, generative models, such as GANs, have traditionally been trained from scratch in an unsupervised manner. Can the collective \"knowledge\" from a large bank of pretrained vision models be leveraged to improve GAN training? If so, with so many models to choose from, which one(s) should be selected, and in what manner are they most effective? We find that pretrained computer vision models can significantly improve performance when used in an ensemble of discriminators. Notably, the particular subset of selected models greatly affects performance. We propose an effective selection mechanism, by probing the linear separability between real and fake samples in pretrained model embeddings, choosing the most accurate model, and progressively adding it to the discriminator ensemble. Interestingly, our method can improve GAN training in both limited data and large-scale settings. Given only 10k training samples, our FID on LSUN Cat matches the StyleGAN2 trained on 1.6M images. On the full dataset, our method improves FID by 1.5x to 2x on cat, church, and horse categories of LSUN.","url_abs":"https://arxiv.org/abs/2112.09130v3","url_pdf":"https://arxiv.org/pdf/2112.09130v3.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":"ensembling-off-the-shelf-models-for-gan","repo_url":"https://github.com/nupurkmr9/vision-aided-gan","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"image-generation","task_name":"Image Generation"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"path-length-regularization","method_name":"Path Length Regularization"},{"method_slug":"r1-regularization","method_name":"R1 Regularization"},{"method_slug":"vision-aided-gan","method_name":"Vision-aided GAN"},{"method_slug":"weight-demodulation","method_name":"Weight Demodulation"}],"datasets_introduced":[],"methods_introduced":[{"slug":"vision-aided-gan","name":"Vision-aided GAN","full_name":"Vision-aided GAN"}],"results":[{"leaderboard":"/sota/image-generation-on-afhq-cat","task":"Image Generation","dataset":"AFHQ Cat","model":"Vision-aided GAN","rank_in_archive_order":1,"of":8,"metrics":{"FID":"2.44","clean-FID":"2.51 ± .02","clean-KID":"0.46 ± .03"},"uses_additional_data":false},{"leaderboard":"/sota/image-generation-on-afhq-dog","task":"Image Generation","dataset":"AFHQ Dog","model":"Vision-aided GAN","rank_in_archive_order":2,"of":6,"metrics":{"FID":"4.60","clean-FID":"4.73 ± .02","clean-KID":"0.38 ± .01"},"uses_additional_data":false},{"leaderboard":"/sota/image-generation-on-afhq-wild","task":"Image Generation","dataset":"AFHQ Wild","model":"Vision-aided GAN","rank_in_archive_order":1,"of":5,"metrics":{"FID":"2.25","clean-FID":"2.35 ± .02","clean-KID":"0.38 ± .02"},"uses_additional_data":false},{"leaderboard":"/sota/image-generation-on-lsun-cat-256-x-256","task":"Image Generation","dataset":"LSUN Cat 256 x 256","model":"Vision-aided GAN","rank_in_archive_order":1,"of":8,"metrics":{"Clean-FID (trainfull)":"3.98 ± 0.03","FID":"3.87"},"uses_additional_data":false},{"leaderboard":"/sota/image-generation-on-lsun-churches-256-x-256","task":"Image Generation","dataset":"LSUN Churches 256 x 256","model":"Vision-aided GAN","rank_in_archive_order":2,"of":27,"metrics":{"Clean-FID (trainfull)":"1.72 ± 0.01","FID":"1.72"},"uses_additional_data":false},{"leaderboard":"/sota/image-generation-on-lsun-horse-256-x-256","task":"Image Generation","dataset":"LSUN Horse 256 x 256","model":"Vision-aided GAN","rank_in_archive_order":1,"of":6,"metrics":{"Clean-FID (trainfull)":"2.11","FID":"2.15"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2112.09130","atlas_url":"https://app.syntology.ai/?focus=2112.09130","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2112.09130"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/nupurkmr9/vision-aided-gan","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran_fixture":1,"ran_draft_wrong":5,"ran":6,"unverified":3},"by_repo_kind":{"official":{"samples":15,"ran":12,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"144d10b49baeb8a6","entry":"window_partition","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/swintaskspecific.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/swintaskspecific.py","link_basis":"harvester_set","language":"python","status":"ran_fixture","verification_level":2,"contract_check":"RAISES","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"144d10b49baeb8a6"}},{"code_sha256_prefix":"fac5364e2f53c6db","entry":"conv3x3","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/resnet.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/resnet.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"fac5364e2f53c6db"}},{"code_sha256_prefix":"bb6429dd2e9f16d3","entry":"convrelu","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/face_normals.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/face_normals.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"bb6429dd2e9f16d3"}},{"code_sha256_prefix":"80a82d5149279cdf","entry":"get_interp_size","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/face_parsing.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/face_parsing.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"80a82d5149279cdf"}},{"code_sha256_prefix":"50aa4a59fe718c61","entry":"get_pad_layer","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/blurpool.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/blurpool.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"50aa4a59fe718c61"}},{"code_sha256_prefix":"176a539246e68879","entry":"get_pad_layer_1d","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/blurpool.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/blurpool.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"176a539246e68879"}},{"code_sha256_prefix":"3e87c5d93ee4df83","entry":"interp","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/face_parsing.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/face_parsing.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"3e87c5d93ee4df83"}},{"code_sha256_prefix":"011230b2b9b8fb6f","entry":"rand_brightness","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/DiffAugment_pytorch.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/DiffAugment_pytorch.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"011230b2b9b8fb6f"}},{"code_sha256_prefix":"5b0d8787e670fc63","entry":"rand_saturation","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/DiffAugment_pytorch.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/DiffAugment_pytorch.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"5b0d8787e670fc63"}},{"code_sha256_prefix":"46b2f3cc882fb768","entry":"resnet18","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/resnet.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/resnet.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"46b2f3cc882fb768"}},{"code_sha256_prefix":"edd87b551d6f38c9","entry":"resnet34","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/resnet.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/resnet.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"edd87b551d6f38c9"}},{"code_sha256_prefix":"61bf152e6a42a184","entry":"window_reverse","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/swintaskspecific.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/swintaskspecific.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"61bf152e6a42a184"}},{"code_sha256_prefix":"1d566d068b0ed0af","entry":"DiffAugment","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/DiffAugment_pytorch.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/DiffAugment_pytorch.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"1d566d068b0ed0af"}},{"code_sha256_prefix":"e475794319393649","entry":"get_upsampling_weight","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/face_parsing.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/face_parsing.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"e475794319393649"}},{"code_sha256_prefix":"4d45510ba267046e","entry":"open_url","repo":"nupurkmr9/vision-aided-gan","repo_kind":"official","path":"vision_aided_loss/utils_.py","file_url":"https://github.com/nupurkmr9/vision-aided-gan/blob/HEAD/vision_aided_loss/utils_.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"4d45510ba267046e"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}