{"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/the-unreasonable-effectiveness-of-deep","title":"The Unreasonable Effectiveness of Deep Features as a Perceptual Metric","arxiv_id":"1801.03924","date":"2018-01-11","proceeding":"CVPR 2018 6","authors":["Richard Zhang","Phillip Isola","Alexei A. Efros","Eli Shechtman","Oliver Wang"],"abstract":"While it is nearly effortless for humans to quickly assess the perceptual\nsimilarity between two images, the underlying processes are thought to be quite\ncomplex. Despite this, the most widely used perceptual metrics today, such as\nPSNR and SSIM, are simple, shallow functions, and fail to account for many\nnuances of human perception. Recently, the deep learning community has found\nthat features of the VGG network trained on ImageNet classification has been\nremarkably useful as a training loss for image synthesis. But how perceptual\nare these so-called \"perceptual losses\"? What elements are critical for their\nsuccess? To answer these questions, we introduce a new dataset of human\nperceptual similarity judgments. We systematically evaluate deep features\nacross different architectures and tasks and compare them with classic metrics.\nWe find that deep features outperform all previous metrics by large margins on\nour dataset. More surprisingly, this result is not restricted to\nImageNet-trained VGG features, but holds across different deep architectures\nand levels of supervision (supervised, self-supervised, or even unsupervised).\nOur results suggest that perceptual similarity is an emergent property shared\nacross deep visual representations.","url_abs":"http://arxiv.org/abs/1801.03924v2","url_pdf":"http://arxiv.org/pdf/1801.03924v2.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":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/richzhang/PerceptualSimilarity","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/EndyWon/Deep-Feature-Perturbation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/Image-X-Institute/lpips_torch2tf","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/Meghraj-Webllisto/stylegan","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/Mohanned-Elkholy/BigGan-model-inversion-with-image-transformations","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/Puzer/stylegan-encoder","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/RudreshVeerkhare/StyleGan","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/SUPERSHOPxyz/stylegan3-gradient","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/ariel415el/PerceptualLossGLO-Pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/ashutosh1919/FaceGenerationStyleGAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/ayushgupta9198/gan","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/ayushgupta9198/stylegan","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/bytedance/LatentSync","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/cassava-math-ubb/experiments","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/isaacschaal/SG_training","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/jooyae/NVIDIA_STYLEGAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/khurram702/StyleBasedGAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/kozistr/gan-metrics","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/pbaylies/stylegan-encoder","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/ryanxingql/image-quality-assessment-toolbox","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/stefkim/stylegan-batik","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/woctezuma/stylegan2-projecting-images","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/zzz2010/starganv2_paddle","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"the-unreasonable-effectiveness-of-deep","repo_url":"https://github.com/S-aiueo32/lpips-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"image-quality-assessment","task_name":"Image Quality Assessment"},{"task_slug":"ssim","task_name":"SSIM"},{"task_slug":"video-quality-assessment","task_name":"Video Quality Assessment"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[{"slug":"perceptual-similarity","name":"Perceptual Similarity","full_name":""}],"methods_introduced":[],"results":[{"leaderboard":"/sota/video-quality-assessment-on-msu-video-quality-1","task":"Video Quality Assessment","dataset":"MSU FR VQA Database","model":"LPIPS","rank_in_archive_order":19,"of":20,"metrics":{"KLCC":"0.5846","PLCC":"0.8128","SRCC":"0.7538"},"uses_additional_data":false},{"leaderboard":"/sota/video-quality-assessment-on-msu-sr-qa-dataset","task":"Video Quality Assessment","dataset":"MSU SR-QA Dataset","model":"LPIPS (Alex)","rank_in_archive_order":28,"of":60,"metrics":{"KLCC":"0.43158","PLCC":"0.52385","SROCC":"0.54461","Type":"FR"},"uses_additional_data":false},{"leaderboard":"/sota/video-quality-assessment-on-msu-sr-qa-dataset","task":"Video Quality Assessment","dataset":"MSU SR-QA Dataset","model":"LPIPS (VGG)","rank_in_archive_order":32,"of":60,"metrics":{"KLCC":"0.41471","PLCC":"0.52820","SROCC":"0.52868","Type":"FR"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1801.03924","atlas_url":"https://app.syntology.ai/?focus=1801.03924","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1801.03924"}},"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":"deterministic:regex_extraction","url":"https://github.com/richzhang/PerceptualSimilarity","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/RudreshVeerkhare/StyleGan","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/pbaylies/stylegan-encoder","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ashutosh1919/FaceGenerationStyleGAN","reach":{"status":"ok","spdx":"NOASSERTION"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/Image-X-Institute/lpips_torch2tf","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/khurram702/StyleBasedGAN","reach":{"status":"ok","spdx":"NOASSERTION"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ayushgupta9198/gan","reach":{"status":"ok","spdx":"NOASSERTION"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/woctezuma/stylegan2-projecting-images","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/EndyWon/Deep-Feature-Perturbation","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/S-aiueo32/lpips-pytorch","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/Puzer/stylegan-encoder","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/bytedance/LatentSync","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ariel415el/PerceptualLossGLO-Pytorch","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/zzz2010/starganv2_paddle","reach":{"status":"ok","spdx":"NOASSERTION"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/stefkim/stylegan-batik","reach":{"status":"ok","spdx":"NOASSERTION"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ayushgupta9198/stylegan","reach":{"status":"ok","spdx":"NOASSERTION"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/Meghraj-Webllisto/stylegan","reach":{"status":"ok","spdx":"NOASSERTION"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/kozistr/gan-metrics","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/jooyae/NVIDIA_STYLEGAN","reach":{"status":"ok","spdx":"NOASSERTION"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/cassava-math-ubb/experiments","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/isaacschaal/SG_training","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/SUPERSHOPxyz/stylegan3-gradient","reach":{"status":"ok","spdx":"NOASSERTION"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/Mohanned-Elkholy/BigGan-model-inversion-with-image-transformations","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ryanxingql/image-quality-assessment-toolbox","reach":{"status":"unanswered"}}],"summary":{"ran_draft_wrong":1,"ran_fixture":1},"by_repo_kind":{"official":{"samples":2,"ran":2,"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":"29e5e72bcd006dcd","entry":"spatial_average","repo":"richzhang/PerceptualSimilarity","repo_kind":"official","path":"lpips/lpips.py","file_url":"https://github.com/richzhang/PerceptualSimilarity/blob/HEAD/lpips/lpips.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"BSD-2-Clause","inline_ok":true,"mcp_get_code":{"code_sha256":"29e5e72bcd006dcd"}},{"code_sha256_prefix":"8cafd18b57a4d171","entry":"upsample","repo":"richzhang/PerceptualSimilarity","repo_kind":"official","path":"lpips/lpips.py","file_url":"https://github.com/richzhang/PerceptualSimilarity/blob/HEAD/lpips/lpips.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"invariant","behaviour_fingerprint":false,"licence":"BSD-2-Clause","inline_ok":true,"mcp_get_code":{"code_sha256":"8cafd18b57a4d171"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}