publications

2026

1. CVPR

   LiteCrackSeg: A Hybrid Architecture for Efficient Crack Segmentation on Edge Devices

   Aeyan Ashraf and collaborators

   2026

   Submitted to CVPR 2026

   Abs Bib

   A 0.55M-parameter, 3.40 GFLOPs hybrid CNN-Transformer model for edge-device crack segmentation, combining depthwise separable convolutions, LoRA-enhanced MobileViT blocks, and a Dynamic Edge Enhancement Module for robust thin-crack detection.

   @article{ashraf2026litecrackseg,
     title = {LiteCrackSeg: A Hybrid Architecture for Efficient Crack Segmentation on Edge Devices},
     author = {Ashraf, Aeyan and collaborators},
     year = {2026},
     note = {Submitted to CVPR 2026},
   }

2025

1. EMNLP

   FINHOP: Benchmarking Retrieval-Augmented Generation for Multi-Hop Questions on Long Financial Documents

   Aeyan Ashraf and collaborators

   2025

   Manuscript under preparation

   Abs Bib

   Presents a grounded evaluation suite for multi-hop financial QA, covering ingestion, retrieval, and reasoning quality for SEC-scale corpora.

   @article{ashraf2025finhop,
     title = {FINHOP: Benchmarking Retrieval-Augmented Generation for Multi-Hop Questions on Long Financial Documents},
     author = {Ashraf, Aeyan and collaborators},
     year = {2025},
     note = {Manuscript under preparation},
   }

2023

1. APSIPA ASC

   A Vision Transformer-Based Approach to Bearing Fault Classification via Vibration Signals

   Aeyan Ashraf and collaborators

   In APSIPA Annual Summit and Conference, 2023

   Abs Bib

   Demonstrates ViT-driven diagnostics that improve industrial bearing fault detection accuracy on vibration datasets.

   @inproceedings{ashraf2023bearing,
     title = {A Vision Transformer-Based Approach to Bearing Fault Classification via Vibration Signals},
     author = {Ashraf, Aeyan and collaborators},
     booktitle = {APSIPA Annual Summit and Conference},
     year = {2023},
     url = {https://link.springer.com/chapter/10.1007/978-981-97-6352-8_59}
   }

2. SIGMAA

   Zea Mays Leaf Disease Classification using Swin-Transformer

   Aeyan Ashraf and collaborators

   In SIGMAA, 2023

   Abs Bib

   Uses Swin-Transformer pipelines for crop-disease identification, enabling earlier agronomy interventions.

   @inproceedings{ashraf2023zea,
     title = {Zea Mays Leaf Disease Classification using Swin-Transformer},
     author = {Ashraf, Aeyan and collaborators},
     booktitle = {SIGMAA},
     year = {2023},
     url = {https://ieeexplore.ieee.org/abstract/document/9980013}
   }