Papers › Detecting Twenty-thousand Classes using Image-level Supervision

Detecting Twenty-thousand Classes using Image-level Supervision

7 Jan 2022arXiv:2201.02605archive 2025-07-28

Xingyi Zhou, Rohit Girdhar, Armand Joulin, Philipp Krähenbühl, Ishan Misra

Current object detectors are limited in vocabulary size due to the small scale of detection datasets. Image classifiers, on the other hand, reason about much larger vocabularies, as their datasets are larger and easier to collect. We propose Detic, which simply trains the classifiers of a detector on image classification data and thus expands the vocabulary of detectors to tens of thousands of concepts. Unlike prior work, Detic does not need complex assignment schemes to assign image labels to boxes based on model predictions, making it much easier to implement and compatible with a range of detection architectures and backbones. Our results show that Detic yields excellent detectors even for classes without box annotations. It outperforms prior work on both open-vocabulary and long-tail detection benchmarks. Detic provides a gain of 2.4 mAP for all classes and 8.3 mAP for novel classes on the open-vocabulary LVIS benchmark. On the standard LVIS benchmark, Detic obtains 41.7 mAP when evaluated on all classes, or only rare classes, hence closing the gap in performance for object categories with few samples. For the first time, we train a detector with all the twenty-one-thousand classes of the ImageNet dataset and show that it generalizes to new datasets without finetuning. Code is available at \url{https://github.com/facebookresearch/Detic}.

PaperPDFCode

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

Code

facebookresearch/Detic officialmentioned in papermentioned on GitHubpytorchApache-2.0 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

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Cross-Domain Few-Shot Object DetectionImage ClassificationOpen Vocabulary Object Detectionimage-classification

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Cross-Domain Few-Shot Object Detection Artaxor Detic-FT mAP 12.0 #14 of 16 Archive leaderboard report
Cross-Domain Few-Shot Object Detection Artaxor Detic(w/o FT) mAP 0.6 #16 of 16 Archive leaderboard report
Cross-Domain Few-Shot Object Detection Clipark1k Detic-FT mAP 22.3 #8 of 10 Archive leaderboard report
Cross-Domain Few-Shot Object Detection Clipark1k Detic(w/o FT) mAP 11.4 #9 of 10 Archive leaderboard report
Cross-Domain Few-Shot Object Detection DIOR Detic-FT mAP 15.4 #12 of 15 Archive leaderboard report
Cross-Domain Few-Shot Object Detection DIOR Detic(w/o FT) mAP 0.1 #15 of 15 Archive leaderboard report
Cross-Domain Few-Shot Object Detection DeepFish Detic-FT mAP 17.9 #7 of 10 Archive leaderboard report
Cross-Domain Few-Shot Object Detection DeepFish Detic(w/o FT) mAP 0.9 #10 of 10 Archive leaderboard report
Cross-Domain Few-Shot Object Detection NEU-DET Detic-FT mAP 16.8 #3 of 10 Archive leaderboard report
Cross-Domain Few-Shot Object Detection NEU-DET Detic(w/o FT) mAP 0.0 #10 of 10 Archive leaderboard report
Cross-Domain Few-Shot Object Detection UODD Detic-FT mAP 16.8 #5 of 16 Archive leaderboard report
Cross-Domain Few-Shot Object Detection UODD Detic(w/o FT) mAP 0.0 #16 of 16 Archive leaderboard report
Open Vocabulary Object Detection LVIS v1.0 Detic AP novel-LVIS base training 17.8 #25 of 28 Archive leaderboard report
Open Vocabulary Object Detection MSCOCO Detic AP 0.5 27.8 #28 of 32 Archive leaderboard report
Open Vocabulary Object Detection OpenImages-v4 Detic AP 0.5 42.2 #2 of 2 Archive leaderboard report
Open Vocabulary Object Detection OpenImages-v4 Detic mask AP50 42.2 #2 of 2 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