Emre Akbas

Links to the papers, code, and short summaries will be shared soon.

3. Explaining CLIP Zero-Shot Predictions Through Concepts.
✍ Onat Özdemir, Anders Christensen, Stephan Alaniz, Zeynep Akata, Emre Akbas.
(Collaboration between @helmholtzmunich.bsky.social and Middle East Technical University)

2. Rethinking Concept Bottleneck Models: From Pitfalls to Solutions.
✍ Merve Taplı, Quentin Bouniot, Wolfgang Stammer, Zeynep Akata, Emre Akbas.
(Collaboration between @helmholtzmunich.bsky.social and Middle East Technical University)

We will present the following papers at hashtag#CVPR 2026 in Denver, CO:

1. MatchED: Crisp Edge Detection Using End-to-End, Matching-Based Supervision.
✍ Bedrettin Çetinkaya, Sinan Kalkan, Emre Akbas.

🔗 Course webpage: user.ceng.metu.edu.tr/~emre/Fall20...

The material may be useful for graduate students and researchers interested in deep learning fundamentals.

Deep Learning - METU CENG 501 - Fall 2025 - YouTube Lecture recordings of CENG501 Deep Learning at METU in Fall 2025. Slides, colab notebooks and recommended readings can be found at https://user.ceng.metu.edu...

I am sharing the lecture recordings of CENG501 Deep Learning.

🎥 Lecture playlist: www.youtube.com/playlist?lis... (They are slides + audio only -- no instructor acting 🙂).

The course page includes slides, Colab notebooks, and recommended readings:

DeepKin: Predicting Relatedness From Low‐Coverage Genomes and Palaeogenomes With Convolutional Neural Networks DeepKin is a novel tool designed to predict relatedness from genomic data using convolutional neural networks (CNNs). Traditional methods for estimating relatedness often struggle when genomic data i...

Our genetic kinship estimation tool “DeepKin” is now available! Our neural network models trained on simulated data work effectively on real ancient data from diverse backgrounds and often outperform available tools. @compevohumang.bsky.social onlinelibrary.wiley.com/doi/10.1111/...

Postdoc fellowship opportunities at METU: enrichtogether.metu.edu.tr/faqs/ for 24 months, good compensation.

NeurIPS participation in Europe We seek to understand if there is interest in being able to attend NeurIPS in Europe, i.e. without travelling to San Diego, US. In the following, assume that it is possible to present accepted papers ...

Would you present your next NeurIPS paper in Europe instead of traveling to San Diego (US) if this was an option? Søren Hauberg (DTU) and I would love to hear the answer through this poll: (1/6)

The companion website is open to all at 384book.net.

Book can be found at www.wiley.com/en-us/Signal...

Wiley did an excellent job creating the e-book version. See the attached video for an excerpt! (The ebook is hosted at vitalsource.com)

Over the years of teaching this course with Prof. Fatos Yarman Vural, we have gradually enriched and transformed her handwritten lecture notes into a textbook. It has been a long and challenging but fun project!

🚀 New Textbook Announcement! 📖

Dear fellow academics, if you're teaching an undergraduate Signals and Systems course, consider using our new textbook: "Signals and Systems: Theory and Practical Explorations with Python".

Dear @csprofkgd.bsky.social , I'd love to be included if there is space. Thank you!

A Ranking-based, Balanced Loss Function Unifying Classification and Localisation in Object Detection We propose average Localisation-Recall-Precision (aLRP), a unified, bounded, balanced and ranking-based loss function for both classification and localisation tasks in object detection. aLRP extends t...

When you want to directly optimize the evaluation measure, e.g. in object detection, you face non-differentiable or zero-gradient components. Then, you manipulate the gradient, see: aLRP loss arxiv.org/abs/2009.13592 and RS loss: arxiv.org/abs/2107.11669.