TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/Counterfactual Attention Learning for Fine-Grained Visual ...

Counterfactual Attention Learning for Fine-Grained Visual Categorization and Re-identification

Yongming Rao, Guangyi Chen, Jiwen Lu, Jie zhou

2021-08-19ICCV 2021 10Few-Shot LearningVehicle Re-IdentificationFine-Grained Visual CategorizationFine-Grained Image RecognitionMitigating Contextual BiasCausal InferencePerson Re-IdentificationFine-Grained Visual RecognitionFine-Grained Image ClassificationImage Categorization
PaperPDFCode(official)

Abstract

Attention mechanism has demonstrated great potential in fine-grained visual recognition tasks. In this paper, we present a counterfactual attention learning method to learn more effective attention based on causal inference. Unlike most existing methods that learn visual attention based on conventional likelihood, we propose to learn the attention with counterfactual causality, which provides a tool to measure the attention quality and a powerful supervisory signal to guide the learning process. Specifically, we analyze the effect of the learned visual attention on network prediction through counterfactual intervention and maximize the effect to encourage the network to learn more useful attention for fine-grained image recognition. Empirically, we evaluate our method on a wide range of fine-grained recognition tasks where attention plays a crucial role, including fine-grained image categorization, person re-identification, and vehicle re-identification. The consistent improvement on all benchmarks demonstrates the effectiveness of our method. Code is available at https://github.com/raoyongming/CAL

Results

TaskDatasetMetricValueModel
Person Re-IdentificationMSMT17Rank-184.2CAL(ResNet50)
Person Re-IdentificationMSMT17mAP64CAL(ResNet50)
Person Re-IdentificationMarket-1501Rank-195.5CAL
Person Re-IdentificationMarket-1501mAP89.5CAL
Person Re-IdentificationDukeMTMC-reIDRank-190CAL
Person Re-IdentificationDukeMTMC-reIDmAP80.5CAL
Few-Shot LearningStanford Cars12-shot Accuracy82.9CAL
Few-Shot LearningStanford Cars16-shot Accuracy88.9CAL
Few-Shot LearningStanford Cars4-shot Accuracy42.2CAL
Few-Shot LearningStanford Cars8-shot Accuracy71.8CAL
Few-Shot LearningFGVC Aircraft12-shot Accuracy67.6CAL
Few-Shot LearningFGVC Aircraft16-shot Accuracy74.3CAL
Few-Shot LearningFGVC Aircraft4-shot Accuracy35.2CAL
Few-Shot LearningFGVC Aircraft8-shot Accuracy55.4CAL
Few-Shot LearningFGVC AircraftHarmonic mean35.2CAL
Few-Shot LearningDTD12-shot Accuracy54.6CAL
Few-Shot LearningDTD16-shot Accuracy57.4CAL
Few-Shot LearningDTD4-shot Accuracy40.9CAL
Few-Shot LearningDTD8-shot Accuracy50.4CAL
Image ClassificationFGVC AircraftAccuracy94.2CAL
Image ClassificationCUB-200-2011Accuracy90.6CAL
Intelligent SurveillanceVehicleID LargeRank-175.1CAL
Intelligent SurveillanceVehicleID LargemAP80.9CAL
Intelligent SurveillanceVehicleID MediumRank-178.2CAL
Intelligent SurveillanceVehicleID MediummAP83.8CAL
Intelligent SurveillanceVeRi-776Rank-195.4CAL
Intelligent SurveillanceVeRi-776Rank597.9CAL
Intelligent SurveillanceVeRi-776mAP74.3CAL
Intelligent SurveillanceVehicleID SmallRank-182.5CAL
Intelligent SurveillanceVehicleID SmallmAP87.8CAL
Fine-Grained Image ClassificationFGVC AircraftAccuracy94.2CAL
Fine-Grained Image ClassificationCUB-200-2011Accuracy90.6CAL
Meta-LearningStanford Cars12-shot Accuracy82.9CAL
Meta-LearningStanford Cars16-shot Accuracy88.9CAL
Meta-LearningStanford Cars4-shot Accuracy42.2CAL
Meta-LearningStanford Cars8-shot Accuracy71.8CAL
Meta-LearningFGVC Aircraft12-shot Accuracy67.6CAL
Meta-LearningFGVC Aircraft16-shot Accuracy74.3CAL
Meta-LearningFGVC Aircraft4-shot Accuracy35.2CAL
Meta-LearningFGVC Aircraft8-shot Accuracy55.4CAL
Meta-LearningFGVC AircraftHarmonic mean35.2CAL
Meta-LearningDTD12-shot Accuracy54.6CAL
Meta-LearningDTD16-shot Accuracy57.4CAL
Meta-LearningDTD4-shot Accuracy40.9CAL
Meta-LearningDTD8-shot Accuracy50.4CAL
Vehicle Re-IdentificationVehicleID LargeRank-175.1CAL
Vehicle Re-IdentificationVehicleID LargemAP80.9CAL
Vehicle Re-IdentificationVehicleID MediumRank-178.2CAL
Vehicle Re-IdentificationVehicleID MediummAP83.8CAL
Vehicle Re-IdentificationVeRi-776Rank-195.4CAL
Vehicle Re-IdentificationVeRi-776Rank597.9CAL
Vehicle Re-IdentificationVeRi-776mAP74.3CAL
Vehicle Re-IdentificationVehicleID SmallRank-182.5CAL
Vehicle Re-IdentificationVehicleID SmallmAP87.8CAL
ClassificationFGVC AircraftOOD Accuracy (%)25.1CAL + ALIA
ClassificationFGVC AircraftTop-1 Accuracy (%)71.8CAL + ALIA
ClassificationFGVC AircraftOOD Accuracy (%)10.2CAL
ClassificationFGVC AircraftTop-1 Accuracy (%)71CAL

Related Papers

GLAD: Generalizable Tuning for Vision-Language Models2025-07-17Weakly Supervised Visible-Infrared Person Re-Identification via Heterogeneous Expert Collaborative Consistency Learning2025-07-17WhoFi: Deep Person Re-Identification via Wi-Fi Channel Signal Encoding2025-07-17Try Harder: Hard Sample Generation and Learning for Clothes-Changing Person Re-ID2025-07-15Mind the Gap: Bridging Occlusion in Gait Recognition via Residual Gap Correction2025-07-15Doodle Your Keypoints: Sketch-Based Few-Shot Keypoint Detection2025-07-10An Enhanced Privacy-preserving Federated Few-shot Learning Framework for Respiratory Disease Diagnosis2025-07-10KeyRe-ID: Keypoint-Guided Person Re-Identification using Part-Aware Representation in Videos2025-07-10