</> Code with Papers by Syntology

Papers › Exploring the Limits of Deep Image Clustering using Pretrained Models

Exploring the Limits of Deep Image Clustering using Pretrained Models

31 Mar 2023arXiv:2303.17896archive 2025-07-28

Nikolas Adaloglou, Felix Michels, Hamza Kalisch, Markus Kollmann

We present a general methodology that learns to classify images without labels by leveraging pretrained feature extractors. Our approach involves self-distillation training of clustering heads based on the fact that nearest neighbours in the pretrained feature space are likely to share the same label. We propose a novel objective that learns associations between image features by introducing a variant of pointwise mutual information together with instance weighting. We demonstrate that the proposed objective is able to attenuate the effect of false positive pairs while efficiently exploiting the structure in the pretrained feature space. As a result, we improve the clustering accuracy over k-means on $17$ different pretrained models by $6.1$\% and $12.2$\% on ImageNet and CIFAR100, respectively. Finally, using self-supervised vision transformers, we achieve a clustering accuracy of $61.6$\% on ImageNet. The code is available at https://github.com/HHU-MMBS/TEMI-official-BMVC2023.

PaperPDFCodeCode Syntology ran

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, Syntology's MCP tool lists every function and class Syntology harvested from this paper and whether it ran (how to connect): get_harvested_code_for_paper(arxiv_id="2303.17896")

Code

Syntology Ran 4 of 4 code samples harvested from 1 repository linked to this paper; 0 have no recorded run. Of those that ran: 3 ran · our draft was wrong; 1 ran · fixture could not drive it.

By repository: official repository: 4 samples from 1 repository, 4 ran. The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

HHU-MMBS/TEMI-official-BMVC2023 officialmentioned in papermentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

4 samples harvested; 4 ran; 0 honoured the contract we drafted; 0 have no recorded run. Read from Syntology's graph 2026-09-24; that is when this build read the record, not when the samples ran.

3ran · our draft was wrong
1ran · fixture could not drive it

Licence: 0 of the 4 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from HHU-MMBS/TEMI-official-BMVC2023. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself; “our draft was wrong” and “fixture could not drive it” are failures of Syntology's instrument, not of the code.

Each sample ends with its code_sha256, Syntology's identity for that exact code. An agent fetches the stored sample with Syntology's MCP tool get_code(code_sha256="…") (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: 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. Samples from a repository marked official are listed first. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of four reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; or the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar). Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

compute_embedding HHU-MMBS/TEMI-official-BMVC2023/gen_embeds.py official repository ran · fixture could not drive it Apache-2.0 (permissive) · a4e72aa97d72ef5e · report
compute_neighbors HHU-MMBS/TEMI-official-BMVC2023/gen_embeds.py official repository ran · our draft was wrong fingerprinted Apache-2.0 (permissive) · 709a9a2f23ee145d · report
default_out_dir HHU-MMBS/TEMI-official-BMVC2023/train_main.py official repository ran · our draft was wrong fingerprinted Apache-2.0 (permissive) · 79b303c122ae61f6 · report
get_outpath HHU-MMBS/TEMI-official-BMVC2023/gen_embeds.py official repository ran · our draft was wrong Apache-2.0 (permissive) · 5914e91451bf1c5e · report

Tasks

ClusteringImage Clustering

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Clustering CIFAR-10 TEMI CLIP ViT-L (openai) ARI 0.932 #4 of 40 Archive leaderboard report
Image Clustering CIFAR-10 TEMI CLIP ViT-L (openai) Accuracy 0.969 #4 of 40 Archive leaderboard report
Image Clustering CIFAR-10 TEMI CLIP ViT-L (openai) Backbone ViT-L #4 of 40 Archive leaderboard report
Image Clustering CIFAR-10 TEMI CLIP ViT-L (openai) NMI 0.926 #4 of 40 Archive leaderboard report
Image Clustering CIFAR-10 TEMI CLIP ViT-L (openai) Train set Train #4 of 40 Archive leaderboard report
Image Clustering CIFAR-10 TEMI DINO ViT-B ARI 0.885 #40 of 40 Archive leaderboard report
Image Clustering CIFAR-10 TEMI DINO ViT-B Accuracy 0.94.5 #40 of 40 Archive leaderboard report
Image Clustering CIFAR-10 TEMI DINO ViT-B Backbone ViT-B #40 of 40 Archive leaderboard report
Image Clustering CIFAR-10 TEMI DINO ViT-B NMI 0.886 #40 of 40 Archive leaderboard report
Image Clustering CIFAR-10 TEMI DINO ViT-B Train set Train #40 of 40 Archive leaderboard report
Image Clustering CIFAR-100 TEMI CLIP ViT-L (openai) ARI 0.612 #4 of 30 Archive leaderboard report
Image Clustering CIFAR-100 TEMI CLIP ViT-L (openai) Accuracy 0.737 #4 of 30 Archive leaderboard report
Image Clustering CIFAR-100 TEMI CLIP ViT-L (openai) NMI 0.799 #4 of 30 Archive leaderboard report
Image Clustering CIFAR-100 TEMI CLIP ViT-L (openai) Train Set Train #4 of 30 Archive leaderboard report
Image Clustering CIFAR-100 TEMI DINO ViT-B ARI 0.533 #5 of 30 Archive leaderboard report
Image Clustering CIFAR-100 TEMI DINO ViT-B Accuracy 0.671 #5 of 30 Archive leaderboard report
Image Clustering CIFAR-100 TEMI DINO ViT-B NMI 0.769 #5 of 30 Archive leaderboard report
Image Clustering CIFAR-100 TEMI DINO ViT-B Train Set Train #5 of 30 Archive leaderboard report
Image Clustering ImageNet TEMI MSN (ViT-L) ARI 48.4 #5 of 12 Archive leaderboard report
Image Clustering ImageNet TEMI MSN (ViT-L) Accuracy 61.6 #5 of 12 Archive leaderboard report
Image Clustering ImageNet TEMI MSN (ViT-L) NMI 82.5 #5 of 12 Archive leaderboard report
Image Clustering ImageNet TEMI DINO (ViT-B) ARI 45.9 #7 of 12 Archive leaderboard report
Image Clustering ImageNet TEMI DINO (ViT-B) Accuracy 58.0 #7 of 12 Archive leaderboard report
Image Clustering ImageNet TEMI DINO (ViT-B) NMI 81.4 #7 of 12 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) TEMI CLIP ViT-L (openai) ACCURACY 0.8343 #1 of 5 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) TEMI CLIP ViT-L (openai) ARI 0.7581 #1 of 5 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) TEMI CLIP ViT-L (openai) NMI 0.9006 #1 of 5 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) TEMI MSN ViT-L ACCURACY 0.8286 #2 of 5 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) TEMI MSN ViT-L ARI 0.7408 #2 of 5 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) TEMI MSN ViT-L NMI 0.8853 #2 of 5 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) TEMI DINO ViT-B ACCURACY 0.7505 #3 of 5 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) TEMI DINO ViT-B ARI 0.6545 #3 of 5 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) TEMI DINO ViT-B NMI 0.8565 #3 of 5 Archive leaderboard report
Image Clustering ImageNet-200 TEMI CLIP ViT-L (openai) ACCURACY 0.7776 #1 of 5 Archive leaderboard report
Image Clustering ImageNet-200 TEMI CLIP ViT-L (openai) ARI 0.6941 #1 of 5 Archive leaderboard report
Image Clustering ImageNet-200 TEMI CLIP ViT-L (openai) NMI 0.8839 #1 of 5 Archive leaderboard report
Image Clustering ImageNet-200 TEMI MSN ViT-L ACCURACY 0.77.96 #2 of 5 Archive leaderboard report
Image Clustering ImageNet-200 TEMI MSN ViT-L ARI 0.667 #2 of 5 Archive leaderboard report
Image Clustering ImageNet-200 TEMI MSN ViT-L NMI 0.8665 #2 of 5 Archive leaderboard report
Image Clustering ImageNet-200 TEMI DINO ViT-B ACCURACY 0.7312 #3 of 5 Archive leaderboard report
Image Clustering ImageNet-200 TEMI DINO ViT-B ARI 0.6231 #3 of 5 Archive leaderboard report
Image Clustering ImageNet-200 TEMI DINO ViT-B NMI 0.852 #3 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) TEMI CLIP ViT-L (openai) ACCURACY 0.8827 #1 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) TEMI CLIP ViT-L (openai) ARI 0.8272 #1 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) TEMI CLIP ViT-L (openai) NMI 0.9232 #1 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) TEMI MSN ViT-L ACCURACY 0.8487 #2 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) TEMI MSN ViT-L ARI 0.7646 #2 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) TEMI MSN ViT-L NMI 0.8814 #2 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) TEMI DINO ViT-B ACCURACY 0.801 #3 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) TEMI DINO ViT-B ARI 0.7093 #3 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) TEMI DINO ViT-B NMI 0.8610 #3 of 5 Archive leaderboard report
Image Clustering STL-10 TEMI DINO ViT-B ARI 0.968 #2 of 29 Archive leaderboard report
Image Clustering STL-10 TEMI DINO ViT-B Accuracy 0.985 #2 of 29 Archive leaderboard report
Image Clustering STL-10 TEMI DINO ViT-B Backbone ViT-B #2 of 29 Archive leaderboard report
Image Clustering STL-10 TEMI DINO ViT-B NMI 0.965 #2 of 29 Archive leaderboard report
Image Clustering STL-10 TEMI DINO ViT-B Train Split Train #2 of 29 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections