GistNet: a Geometric Structure Transfer Network for Long-Tailed Recognition

Liu, Bo; Li, Haoxiang; Kang, Hao; Hua, Gang; Vasconcelos, Nuno

doi:10.1109/iccv48922.2021.00810

Cited by 35 publications

(11 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…During training, the tail class-specific features are fused with the head class-generic features to generate new features to augment the tail classes. GIST [38] proposes to transfer the geometric information of the feature distribution boundaries of the head classes to the tail classes by increasing the classifier weights of the tail classes. The motivation of the recently proposed CMO [45] is very intuitive, it argues that the images from the head classes have rich backgrounds, so the images from the tail classes can be pasted directly onto the rich background images of the head classes to increase the richness of the tail images.…”

Section: Information Augmentationmentioning

confidence: 99%

“…The imbalance of sample numbers in the dataset gives rise to the challenge of long-tailed visual recognition. Most previous works assume that head classes are always easier to be learned than tail classes, e.g., class re-balancing [8,14,24,34,37,46,52], information augmentation [23,31,35,38,39,44,56,64,67], decoupled training [10,16,29,30,71,76], and ensemble learning [20,36,57,58,61,72,77] have been proposed to improve the performance of tail classes. However, recent studies [3,50] have shown that classification dif- ficulty is not always correlated with the number of samples, e.g., the performance of some tail classes is even higher than that of the head classes.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A multimodal hyper-fusion transformer for remote sensing image classification

Jiao

et al. 2023

Information Fusion

View full text Add to dashboard Cite

Section: Information Augmentationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A multimodal hyper-fusion transformer for remote sensing image classification

Jiao

et al. 2023

Information Fusion

View full text Add to dashboard Cite

“…To verify the effectiveness, we compare the proposed method ResCom against four groups of stateof-the-art methods on common-used long-tailed datasets: Logits modification methods, including LDAM [3], Causal Norm [44], Balanced Softmax [37], and LADE [22]. Architecture modification methods, like BBN [56], ELF [15], ResLT [11], GistNet [33], RIDE [49], and DiVE [20]. Multi-stage methods such as Decouple [27], DisAlign [53], MiSLAS [54], SSD [30].…”

Section: Comparison With State-of-the-artsmentioning

confidence: 99%

Rebalanced Siamese Contrastive Mining for Long-Tailed Recognition

Zhong¹,

Cui²,

Lo³

et al. 2022

Preprint

View full text Add to dashboard Cite

Deep neural networks perform poorly on heavily class-imbalanced datasets. Given the promising performance of contrastive learning, we propose Rebalanced Siamese Contrastive Mining (ResCom) to tackle imbalanced recognition. Based on the mathematical analysis and simulation results, we claim that supervised contrastive learning suffers a dual class-imbalance problem at both the original batch and Siamese batch levels, which is more serious than long-tailed classification learning. In this paper, at the original batch level, we introduce a class-balanced supervised contrastive loss to assign adaptive weights for different classes. At the Siamese batch level, we present a class-balanced queue, which maintains the same number of keys for all classes. Furthermore, we note that the contrastive loss gradient with respect to the contrastive logits can be decoupled into the positives and negatives, and easy positives and easy negatives will make the contrastive gradient vanish. We propose supervised hard positive and negative pairs mining to pick up informative pairs for contrastive computation and improve representation learning. Finally, to approximately maximize the mutual information between the two views, we propose Siamese Balanced Softmax and joint it with the contrastive loss for one-stage training. ResCom outperforms the previous methods by large margins on multiple long-tailed recognition benchmarks. Our code will be made publicly available at: https://github.com/dvlab-research/ResCom. 1

show abstract

“…Class rebalancing [2,3,7,14,17,18,20,23,26,28,33,34,37,39,41,42,45], for example, aims to boost the weight of losses arising from tail classes, thereby pushing the decision boundary away from the tail class and improving the probability of correctly classifying the underlying distribution. Information augmentation [4,8,12,15,19,21,22,24,25,32,36,38,39,43], on the other hand, expands the observed distribution of the tail classes by introducing prior knowledge to facilitate the model learning of the underlying distribution. It is important to note that these methods default to the model being able to learn adequately and fairly at least for the samples in the training domain.…”

Section: Introductionmentioning

confidence: 99%

Orthogonal Uncertainty Representation of Data Manifold for Robust Long-Tailed Learning

Ma,

Jiao,

Liu

et al. 2023

Proceedings of the 31st ACM International Conference on Multimedia

View full text Add to dashboard Cite

In scenarios with long-tailed distributions, the model's ability to identify tail classes is limited due to the under-representation of tail samples. Class rebalancing, information augmentation, and other techniques have been proposed to facilitate models to learn the potential distribution of tail classes. The disadvantage is that these methods generally pursue models with balanced class accuracy on the data manifold, while ignoring the ability of the model to resist interference. By constructing noisy data manifold, we found that the robustness of models trained on unbalanced data has a long-tail phenomenon. That is, even if the class accuracy is balanced on the data domain, it still has bias on the noisy data manifold. However, existing methods cannot effectively mitigate the above phenomenon, which makes the model vulnerable in long-tailed scenarios. In this work, we propose an Orthogonal Uncertainty Representation (OUR) of feature embedding and an end-to-end training strategy to improve the long-tail phenomenon of model robustness. As a general enhancement tool, OUR has excellent compatibility with other methods and does not require additional data generation, ensuring fast and efficient training. Comprehensive evaluations on long-tailed datasets show that our method significantly improves the long-tail phenomenon of robustness, bringing consistent performance gains to other long-tailed learning methods. CCS CONCEPTS• Computing methodologies → Computer vision; Image representations; Supervised learning by classification.

show abstract

GistNet: a Geometric Structure Transfer Network for Long-Tailed Recognition

Cited by 35 publications

References 14 publications

A multimodal hyper-fusion transformer for remote sensing image classification

A multimodal hyper-fusion transformer for remote sensing image classification

Rebalanced Siamese Contrastive Mining for Long-Tailed Recognition

Orthogonal Uncertainty Representation of Data Manifold for Robust Long-Tailed Learning

Contact Info

Product

Resources

About