Revisiting class-incremental learning with pre-trained models: Generalizability and adaptivity are all you need
Class-incremental learning (CIL) aims to adapt to emerging new classes without forgetting
old ones. Traditional CIL models are trained from scratch to continually acquire knowledge …
old ones. Traditional CIL models are trained from scratch to continually acquire knowledge …
Ma-sam: Modality-agnostic sam adaptation for 3d medical image segmentation
Abstract The Segment Anything Model (SAM), a foundation model for general image
segmentation, has demonstrated impressive zero-shot performance across numerous …
segmentation, has demonstrated impressive zero-shot performance across numerous …
Revisiting class-incremental learning with pre-trained models: Generalizability and adaptivity are all you need
Class-incremental learning (CIL) aims to adapt to emerging new classes without forgetting
old ones. Traditional CIL models are trained from scratch to continually acquire knowledge …
old ones. Traditional CIL models are trained from scratch to continually acquire knowledge …
Convolutional bypasses are better vision transformer adapters
The pretrain-then-finetune paradigm has been widely adopted in computer vision. But as the
size of Vision Transformer (ViT) grows exponentially, the full finetuning becomes prohibitive …
size of Vision Transformer (ViT) grows exponentially, the full finetuning becomes prohibitive …
One-for-all: Generalized lora for parameter-efficient fine-tuning
We present Generalized LoRA (GLoRA), an advanced approach for universal parameter-
efficient fine-tuning tasks. Enhancing Low-Rank Adaptation (LoRA), GLoRA employs a …
efficient fine-tuning tasks. Enhancing Low-Rank Adaptation (LoRA), GLoRA employs a …
Revisiting the parameter efficiency of adapters from the perspective of precision redundancy
Current state-of-the-art results in computer vision depend in part on fine-tuning large pre-
trained vision models. However, with the exponential growth of model sizes, the …
trained vision models. However, with the exponential growth of model sizes, the …
Mind the interference: Retaining pre-trained knowledge in parameter efficient continual learning of vision-language models
This study addresses the Domain-Class Incremental Learning problem, a realistic but
challenging continual learning scenario where both the domain distribution and target …
challenging continual learning scenario where both the domain distribution and target …
Few-shot adaptation of multi-modal foundation models: A survey
F Liu, T Zhang, W Dai, C Zhang, W Cai, X Zhou… - Artificial Intelligence …, 2024 - Springer
Abstract Multi-modal (vision-language) models, such as CLIP, are replacing traditional
supervised pre-training models (eg, ImageNet-based pre-training) as the new generation of …
supervised pre-training models (eg, ImageNet-based pre-training) as the new generation of …
Pyra: Parallel yielding re-activation for training-inference efficient task adaptation
Recently, the scale of transformers has grown rapidly, which introduces considerable
challenges in terms of training overhead and inference efficiency in the scope of task …
challenges in terms of training overhead and inference efficiency in the scope of task …