Get Started
Research questionHow can object detectors account for information-density imbalance when category instance counts are balanced?Object detectors can perform unevenly across categories even when those categories have similar numbers of training instances. Differences in the information density of categories may help explain this persistent bias.
Computer Vision
Machine Learning
Latest papersRecent research connected to this question, newest first.Information Density Imbalance in Visual Object DetectionThe source defines and measures information density, examines its relationship with category accuracy and training, and reports improvements to object-detection loss functions on Pascal VOC, COCO-LT, and LVIS. The evidence concerns category bias and overall detection performance in these datasets.research paper · Sep 2, 2026
Related questions
How can infrared-visible object detectors remain reliable when one sensor modality is missing?How should multi-expert models balance diversity and prediction variance for long-tailed visual recognition?How can object detectors trained on uncamouflaged imagery generalize to unseen camouflage with little target-domain data?How can vision systems discover unknown fine-grained classes without capture-device bias distorting cross-device evaluation?
Home
Topics
Search
Library