AI Paper Atlas

Research map / Benchmarks / PASCAL VOC 2012

PASCAL VOC 2012 leaderboard

277 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on PASCAL VOC 2012. Best mIoU: 90.6 (SegNeXt-L, NeurIPS 2022).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

PASCAL VOC 2012 · mIoU

151 papers, each with its best reported number for its own method; higher is better.

#mIoU ↑MethodPaperTopic
190.6SegNeXt-L
MSCAN-L
SegNeXt: Rethinking Convolutional Attention Design for Semantic Segmentation
NeurIPS 2022 · Table: Pascal VOC
CNN features
290.56NeuBoots
DeepLab-v3, ResNet-101
Neural Bootstrapper
NeurIPS 2021 · Table: Semantic segmentation
Recognition and long-tail
390.5Eff-L2 w/ Self-training
EfficientNet-L2 NAS-FPN
Rethinking Pre-training and Self-training
NeurIPS 2020 · Table 6
Open-set and few-shot detection
489DeepLabv3+ (Xception-JFT)Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation
ECCV 2018 · Table 5
CNN features
588.7ZigZagNet
ResNet-101
ZigZagNet: Fusing Top-Down and Bottom-Up Context for Object Segmentation
CVPR 2019 · Table 3
CNN features
688MSCI
ResNet-152
Multi-Scale Context Intertwining for Semantic Segmentation
ECCV 2018 · Table 3
CNN features
787.9DPC
Modified Xception
Searching for Efficient Multi-Scale Architectures for Dense Image Prediction
NeurIPS 2018 · Table: PASCAL VOC 2012 test se
CNN features
887.9ExFuse
ResNeXt131
ExFuse: Enhancing Feature Fusion for Semantic Segmentation
ECCV 2018 · Table 7
CNN features
987.6EfficientFCN (COCO pretrained)
ResNet101
A Holistically-Guided Decoder for Deep Representation Learning With Applications to Semantic Segmentation and Object Detection
TPAMI 2021 · Table (VOC)
CNN features
1087.2CFNet
ResNet-101
Co-Occurrent Features in Semantic Segmentation
CVPR 2019 · Table 5
CNN features
141 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

PASCAL VOC 2012 · mIoU (val)

144 papers, each with its best reported number for its own method; higher is better.

#mIoU ↑MethodPaperTopic
190Eff-L2 w/ Self-training
EfficientNet-L2 NAS-FPN
Rethinking Pre-training and Self-training
NeurIPS 2020 · Table 6
Open-set and few-shot detection
288.4Region Representation (SAM+SLIC)
DINOv2 ViT-L/14
Region-Based Representations Revisited
CVPR 2024 · Table 7
Open-set and few-shot detection
386.6Eff-B7 NAS-FPN with Copy-Paste pre-training
Eff-B7 NAS-FPN
Simple Copy-Paste Is a Strong Data Augmentation Method for Instance Segmentation
CVPR 2021 · Table: PASCAL VOC semantic seg
Open-set and few-shot detection
486.3ModuSegModuSeg: Decoupling Object Discovery and Semantic Retrieval for Training-Free Weakly Supervised Segmentation
ECCV 2026 · Table 1
Weakly supervised detection and segmentation
586ZigZagNet
ResNet-101
ZigZagNet: Fusing Top-Down and Bottom-Up Context for Object Segmentation
CVPR 2019 · Table 3
CNN features
685.1MSCI
ResNet-152
Multi-Scale Context Intertwining for Semantic Segmentation
ECCV 2018 · Table 3
CNN features
784.9DiffuMask S: 60.0k + R: 5.0k finetune
Mask2Former Swin-B
DiffuMask: Synthesizing Images with Pixel-level Annotations for Semantic Segmentation Using Diffusion Models
ICCV 2023 · Table 1
Fast diffusion sampling
884.1FDNet-16s-finetuning-MS
FDNet-16s
Learning Fully Dense Neural Networks for Image Semantic Segmentation
AAAI 2019 · Table: Comparison of different
CNN features
983.1PSP + PNA
PSP
Progressive Neighborhood Aggregation for Semantic Segmentation Refinement
AAAI 2023 · Table 3
CNN features
1083.1Lovász PixIoU (Lv. Pix)
DeepLabv3 ResNet-101
Learning Generalized Intersection Over Union for Dense Pixelwise Prediction
ICML 2021 · Table 3
Prediction and decision losses
134 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

PASCAL VOC 2012 · mAP

71 papers, each with its best reported number for its own method; higher is better.

#mAP ↑MethodPaperTopic
196.1CCD-R101
ResNet101
Contextual Debiasing for Visual Recognition With Causal Mechanisms
CVPR 2022 · Table 3
Recognition and long-tail
295.5ADD-GCN
ResNet-101
Attention-Driven Dynamic Graph Convolutional Network for Multi-Label Image Recognition
ECCV 2020 · Table 3
Recognition and long-tail
395.4SSGRL (Ours, pre-trained and two-scale fusion)
ResNet-101
Learning Semantic-Specific Graph Representation for Multi-Label Image Recognition
ICCV 2019 · Table 3
Recognition and long-tail
494.6ResNet-cut + CSRAResidual Attention: A Simple but Effective Method for Multi-Label Recognition
ICCV 2021 · Table 3
Recognition and long-tail
593.68CMLL
ViT-B/32
Towards Reliable Multi-Label Classification via Conditional Dependency Modeling
ECCV 2026 · Table 1
Recognition and long-tail
693.4WILDCAT
ResNet-101
WILDCAT: Weakly Supervised Learning of Deep ConvNets for Image Classification, Pointwise Localization and Segmentation
CVPR 2017 · Table 2
Weakly supervised detection and segmentation
793.2DSDL
ResNet-101
Deep Semantic Dictionary Learning for Multi-label Image Classification
AAAI 2021 · Table 2
Recognition and long-tail
892.5G-MIML-FCN+BBMIML-FCN+: Multi-Instance Multi-Label Learning via Fully Convolutional Networks With Privileged Information
CVPR 2017 · Classification Results table
CNN features
991.4Components: Estimate + Correct + Adapt (full)Adapting Vision-Language Models from Iconic to Inclusive for Multi-label Recognition Without Labels
IJCV 2026 · Table 3
CLIP and zero-shot
1090.7Fusion
16-layer CNN
Exploit Bounding Box Annotations for Multi-Label Object Recognition
CVPR 2016 · Table: VOC 2012 classification
CNN features
61 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.