HyperAI超神経

プラットフォーム

Pedestrian Attribute Recognition On Pa 100K

評価指標

Accuracy

評価結果

このベンチマークにおける各モデルのパフォーマンス結果

		Paper Title
PATH (Partial FT)	90.8	HumanBench: Towards General Human-centric Perception with Projector Assisted Pretraining
Hulk(Finetune, ViT-L)	88.97	Hulk: A Universal Knowledge Translator for Human-Centric Tasks
Hulk(Finetune, ViT-B)	87.85	Hulk: A Universal Knowledge Translator for Human-Centric Tasks
C2T-Net	87.2	C2T-Net: Channel-Aware Cross-Fused Transformer-Style Networks for Pedestrian Attribute Recognition
SOLIDER	86.38	Beyond Appearance: a Semantic Controllable Self-Supervised Learning Framework for Human-Centric Visual Tasks
UniHCP (finetune)	86.18	UniHCP: A Unified Model for Human-Centric Perceptions
APTM	80.17	Towards Unified Text-based Person Retrieval: A Large-scale Multi-Attribute and Language Search Benchmark
Label2Label	79.23	Label2Label: A Language Modeling Framework for Multi-Attribute Learning
strongBaseline(ours)	78.56	Rethinking of Pedestrian Attribute Recognition: Realistic Datasets with Efficient Method
Attribute-Specific Localization	77.08%	Improving Pedestrian Attribute Recognition With Weakly-Supervised Multi-Scale Attribute-Specific Localization
HP-net	72.19%	HydraPlus-Net: Attentive Deep Features for Pedestrian Analysis

0 of 11 row(s) selected.

Pedestrian Attribute Recognition On Pa 100K | SOTA | HyperAI超神経