SeeThroughSmoke
5,817 pairs · 1280 × 1024 · Hugging Face
A paired surgical smoke and smoke-free image dataset for training and evaluating desmoking models, with 3,097 training pairs and 2,720 test pairs across multiple anatomical regions.
Bibliography
Scholar-indexed work in surgical vision datasets, digital desmoking, high-resolution endoscopy, and biomedical question answering.
Google Scholar profilearXiv 2026 · Preprint
A surgical vision paper on real-time digital desmoking for robot-assisted procedures. It introduces a transformer-based model with a physics-inspired desmoking head, a synthetic paired-data pipeline, and a 5,817-pair real stereo smoke benchmark for evaluating reconstruction quality and downstream perception.
Surgical Vision · Desmoking · Transformer · Dataset
MELBA 2025 · Dataset Paper
A public native-4K surgical endoscopy dataset for robotic-assisted minimally invasive procedures. The dataset provides high-resolution images and clips captured under realistic surgical conditions, supporting super-resolution, smoke removal, instrument detection, depth estimation, segmentation, novel-view synthesis, and VLM research.
4K Endoscopy · Super-Resolution · Medical Imaging · Dataset
CLEF 2025 · Workshop Paper
A biomedical QA system for the BioASQ 2025 Task13b challenge. The work combines dense retrieval, fine-tuned cross-encoders, GPT rerankers, and instruction-tuned LLM prompting to retrieve PubMed evidence and generate exact and ideal answers.
Biomedical QA · RAG · Reranking · LLMs
Appendix A
5,817 pairs · 1280 × 1024 · Hugging Face
A paired surgical smoke and smoke-free image dataset for training and evaluating desmoking models, with 3,097 training pairs and 2,720 test pairs across multiple anatomical regions.
4K Endoscopy · Synapse
A native-4K endoscopic imaging dataset for robotic-assisted surgery research, including high-resolution still images and short surgical video clips.
Biomedical QA · Benchmark · CLEF 2025
The biomedical semantic question answering benchmark used by the Beyond Retrieval paper for PubMed document retrieval, snippet retrieval, and answer generation.