While foundation models have revolutionised computer vision, their effectiveness for sketch understanding remains limited by the unique challenges of abstract, sparse visual inputs. Through systematic analysis, we uncover two fundamental limitations: Stable Diffusion (SD) struggles to extract meaningful features from abstract sketches (unlike its success with photos), and exhibits a pronounced frequency-domain bias that suppresses essential low-frequency components needed for sketch understanding. Rather than costly retraining, we address these limitations by strategically combining SD with CLIP, whose strong semantic understanding naturally compensates for SD’s spatial-frequency biases. By dynamically injecting CLIP features into SD’s denoising process and adaptively aggregating features across semantic levels, our method achieves state-of-the-art performance in sketch retrieval (+3.35%), recognition (+1.06%), segmentation (+29.42%), and correspondence learning (+21.22%), demonstrating the first truly universal sketch feature representation in the era of foundation models.

Polyp segmentation in colonoscopy is crucial for detecting colorectal cancer. However, it is challenging due to variations in the structure, color, and size of polyps, as well as the lack of clear boundaries with surrounding tissues. Traditional segmentation models based on Convolutional Neural Networks (CNNs) struggle to capture detailed patterns and global context, limiting their performance. Vision Transformer (ViT)-based models address some of these issues but have difficulties in capturing local context and lack strong zero-shot generalization. To this end, we propose the Mamba-guided Segment Anything Model (SAM-Mamba††Code, Modes:https://github.com/TapasKumarDuttal/SAM_Mamba_2025) for efficient polyp segmentation. Our approach introduces a Mamba-Prior module in the encoder to bridge the gap between the general pre-trained representation of SAM and polyp-relevant trivial clues. It injects salient cues of polyp images into the SAM image encoder as a domain prior while capturing global dependencies at various scales, leading to more accurate segmentation results. Extensive experiments on five benchmark datasets show that SAM-Mamba outper-forms traditional CNN, ViT, and Adapter-based models in both quantitative and qualitative measures. Additionally, SAM-Mamba demonstrates excellent adaptability to unseen datasets, making it highly suitable for real-time clinical use.

Brain tumor is the most common type of cancer that causes a high mortality rate among individuals of all age groups. Hence, accurate diagnosis of brain tumor and its type at early stages is of utmost importance to preclude its severity and ultimately helps patients for timely and better treatment. The current literature has witnessed the usage of convolutional neural networks (CNN) for brain tumor classification in MR images; however, such traditional CNN methods may fail to identify the minute variations of tumor lesions. This paper proposes an attention module called channel split dual attention (CSDA) coupled with a backbone network to handle this issue. The CSDA explores more detailed and discriminative features using channel splitting and two parallel attention blocks: position attention block (PAB) and channel attention block (CAB). The PAB and CAB are introduced to capture feature dependencies in spatial and channel dimensions. Extensive experiments on a publicly available dataset show that our CDANet significantly improves the state-of-the-art CNN results and obtains higher classification accuracy than existing brain tumor detection methods. The codes and models are available at https://github.com/TapasKumarDutta1/CDANet.