Searcharxiv⌕ Search

arXiv subjects

Jaeyoon Seo

Publications and source records attributed to Jaeyoon Seo.

4 recordsLinked to original sources

Drift-Conditioned Structural Replay for Continual Cardiac Ultrasound Segmentation

Conventional experience replay typically applies fixed preservation constraints to historical exemplars, overlooking variations in how learned representations, predictions, and anatomical structures change during sequential training. To address this limitation, we propose Drift-Conditioned Structural Replay Loss (DCSRL), an exemplar-adaptive supervision mechanism that regulates anatomical preservation according to source-referenced knowledge drift. DCSRL operates within a reliability-guided replay framework that selects informative historical exemplars using anatomical content and boundary characteristics while adaptively regulating replay during training. For each retained exemplar, DCSRL constructs a frozen source-task reference signature and subsequently measures changes in feature representations, class probabilities, boundary responses, and predicted anatomical structures. These complementary drift measurements are aggregated and normalized to generate bounded, exemplar-specific coefficients that dynamically adjust boundary and structural replay losses without modifying the underlying memory-selection policy. We evaluate DCSRL under sequential learning across CAMUS, CardiacNet, and EchoNet-Dynamic, comparing its performance with regularization-, distillation-, and replay-based continual-learning approaches. Experimental findings indicate that DCSRL improves overall final segmentation performance and anatomical boundary preservation, particularly under heterogeneous imaging conditions, although trade-offs remain in knowledge transfer across tasks. These findings suggest that source-referenced multidimensional drift conditioning offers a promising direction for adaptive anatomical knowledge preservation in continual cardiac ultrasound segmentation.

cs.CV↗

StableSketcher: Enhancing Diffusion Model for Pixel-based Sketch Generation via Visual Question Answering Feedback

Although recent advancements in diffusion models have significantly enriched the quality of generated images, challenges remain in synthesizing pixel-based human-drawn sketches, a representative example of abstract expression. To combat these challenges, we propose StableSketcher, a novel framework that empowers diffusion models to generate hand-drawn sketches with high prompt fidelity. Within this framework, we fine-tune the variational autoencoder to optimize latent decoding, enabling it to better capture the characteristics of sketches. In parallel, we integrate a new reward function for reinforcement learning based on visual question answering, which improves text-image alignment and semantic consistency. Extensive experiments demonstrate that StableSketcher generates sketches with improved stylistic fidelity, achieving better alignment with prompts compared to the Stable Diffusion baseline. Additionally, we introduce SketchDUO, to the best of our knowledge, the first dataset comprising instance-level sketches paired with captions and question-answer pairs, thereby addressing the limitations of existing datasets that rely on image-label pairs. Our code and dataset will be made publicly available upon acceptance. Project page: https://zihos.github.io/StableSketcher

cs.CV↗

Before We Trust Them: Decision-Making Failures in Navigation of Foundation Models

High success rates on navigation-related tasks do not necessarily translate into reliable decision making by foundation models. To examine this gap, we evaluate current models on six diagnostic tasks spanning three settings: reasoning under complete spatial information, reasoning under incomplete spatial information, and reasoning under safety-relevant information. Our results show that the current metrics may not capture critical limitations of the models and indicate good performance, underscoring the need for failure-focused analysis to understand model limitations and guide future progress. In a path-planning setting with unknown cells, GPT-5 achieved a high success rate of 93%; Yet, the failed cases exhibit fundamental limitations of the models, e.g., the lack of structural spatial understanding essential for navigation. We also find that newer models are not always more reliable than their predecessors on this end. In reasoning under safety-relevant information, Gemini-2.5 Flash achieved only 67% on the challenging emergency-evacuation task, underperforming Gemini-2.0 Flash, which reached 100% under the same condition. Across all evaluations, models exhibited structural collapse, hallucinated reasoning, constraint violations, and unsafe decisions. These findings show that foundation models still exhibit substantial failures in navigation-related decision making and require fine-grained evaluation before they can be trusted.

cs.AI↗

SEA: Evaluating Sketch Abstraction Efficiency via Element-level Commonsense Visual Question Answering

A sketch is a distilled form of visual abstraction that conveys core concepts through simplified yet purposeful strokes while omitting extraneous detail. Despite its expressive power, quantifying the efficiency of semantic abstraction in sketches remains challenging. Existing evaluation methods that rely on reference images, low-level visual features, or recognition accuracy do not capture abstraction, the defining property of sketches. To address these limitations, we introduce SEA (Sketch Evaluation metric for Abstraction efficiency), a reference-free metric that assesses how economically a sketch represents class-defining visual elements while preserving semantic recognizability. These elements are derived per class from commonsense knowledge about features typically depicted in sketches. SEA leverages a visual question answering model to determine the presence of each element and returns a quantitative score that reflects semantic retention under visual economy. To support this metric, we present CommonSketch, the first semantically annotated sketch dataset, comprising 23,100 human-drawn sketches across 300 classes, each paired with a caption and element-level annotations. Experiments show that SEA aligns closely with human judgments and reliably discriminates levels of abstraction efficiency, while CommonSketch serves as a benchmark providing systematic evaluation of element-level sketch understanding across various vision-language models.

cs.CV↗