Papers › D3Former: Debiased Dual Distilled Transformer for Incremental Learning

D3Former: Debiased Dual Distilled Transformer for Incremental Learning

25 Jul 2022arXiv:2208.00777archive 2025-07-28

Abdelrahman Mohamed, Rushali Grandhe, K J Joseph, Salman Khan, Fahad Khan

In class incremental learning (CIL) setting, groups of classes are introduced to a model in each learning phase. The goal is to learn a unified model performant on all the classes observed so far. Given the recent popularity of Vision Transformers (ViTs) in conventional classification settings, an interesting question is to study their continual learning behaviour. In this work, we develop a Debiased Dual Distilled Transformer for CIL dubbed D³Former. The proposed model leverages a hybrid nested ViT design to ensure data efficiency and scalability to small as well as large datasets. In contrast to a recent ViT based CIL approach, our D³Former does not dynamically expand its architecture when new tasks are learned and remains suitable for a large number of incremental tasks. The improved CIL behaviour of D³Former owes to two fundamental changes to the ViT design. First, we treat the incremental learning as a long-tail classification problem where the majority samples from new classes vastly outnumber the limited exemplars available for old classes. To avoid the bias against the minority old classes, we propose to dynamically adjust logits to emphasize on retaining the representations relevant to old tasks. Second, we propose to preserve the configuration of spatial attention maps as the learning progresses across tasks. This helps in reducing catastrophic forgetting by constraining the model to retain the attention on the most discriminative regions. D³Former obtains favorable results on incremental versions of CIFAR-100, MNIST, SVHN, and ImageNet datasets. Code is available at https://tinyurl.com/d3former

Paper PDF Code

In Syntology View this paper on Syntology: its repositories, every harvested function with whether it ran, its licence and the call to fetch it.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

abdohelmy/D-3Former officialmentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Continual Learning Incremental Learning class-incremental learning

Results from the paper archive 2025-07-28

Task	Dataset	Model	Metric	Value	Rank at snapshot	Leaderboard	Report
Incremental Learning	CIFAR-100 - 50 classes + 10 steps of 5 classes	D3Former	Average Incremental Accuracy	70.94	#4 of 13	Archive leaderboard	report
Incremental Learning	CIFAR-100 - 50 classes + 25 steps of 2 classes	D3Former	Average Incremental Accuracy	68.68	#1 of 5	Archive leaderboard	report
Incremental Learning	CIFAR-100 - 50 classes + 5 steps of 10 classes	D3Former	Average Incremental Accuracy	72.23	#4 of 15	Archive leaderboard	report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Absolute Position Encodings Adam Attention BPE Dense Connections Dropout Label Smoothing Layer Normalization Linear Layer Multi-Head Attention Position-Wise Feed-Forward Layer Residual Connection Softmax Transformer

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections