fatih c. akyon

AI eng. @ultralytics, PhD @metu_odtu.
coauthor of YOLO26, leading YOLO27 multimodal backbone.
10M+ open-source downloads, 8 patents, 1500+ citations.

[fatih c. akyon @fcakyon]

I have been working on this for a very long time, finally it's live on Arxiv!

Introducing the long-awaited @Ultralytics YOLO26 paper: state-of-the-art real-time object detection, instance segmentation, image classification, keypoint detection, oriented bounding boxes detection and open-vocabulary detection models that run on any edge device 🔥

Meet me at @CVPR booth #513 if you have questions about the paper!

It was a lot of fun collaborating with my coauthors @arttestt and others.

Of course, it wouldn't be possible without the ultralytics package maintainers like @onuralpszr and rest of the team.

[fatih c. akyon @fcakyon]

Find me at the @CVPR booth 513 if you want to get some Turkish delight 😋

[Ultralytics @ultralytics]

#CVPR2026 - See you tomorrow! 🚀

We're excited to be exhibiting at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (@CVPR) in Denver!

Stop by our booth to learn about our latest research, including our new Ultralytics YOLO26 paper and experience live demos of cutting-edge vision AI.

📍 Booth 513
📅 June 5 - 7, 2026 | Colorado Convention Center, Denver

Learn more ➡️ https://t.co/UHdvAiJw2Q

[fatih c. akyon @fcakyon]

@onuralpszr

[fatih c. akyon @fcakyon]

1st day at @CVPR was fun 😎

☑️ Presented our paper, SenBen (Sensitive Benchmark), introducing a new dataset with explicit images and labels for vision-language models and a new metric to evaluate VLM's. Also, several tricks that improve the multimodal multitask generation capabilities of autoregressive models 👌
☑️ Published the @Ultralytics YOLO26 paper at Arxiv.
☑️ Joined some sessions with my PhD advisor, Prof. Alptekin Temizel.

Also met with amazing people from the open-source community (everyone is asking about the YOLO26 paper 😂)

[fatih c. akyon @fcakyon]

.@CVPR'da SenBen calismamizi sunarken profesorum ve ben 😅

[Google Gemma @googlegemma]

Meet Gemma 4 12B!

A unified, encoder-free multimodal model designed to bring high-performance intelligence directly to your laptop, and released under an Apache 2.0 license.

Bridging the gap between edge efficiency and advanced reasoning. Here is what’s new with Gemma 4 12B: 👇

[fatih c. akyon @fcakyon]

link to paper: https://t.co/iJgS7bjkhO

[fatih c. akyon @fcakyon]

I have been working on this for a very long time, finally it's live on Arxiv!

Introducing the long-awaited @Ultralytics YOLO26 paper: state-of-the-art real-time object detection, instance segmentation, image classification, keypoint detection, oriented bounding boxes detection and open-vocabulary detection models that run on any edge device 🔥

Meet me at @CVPR booth #513 if you have questions about the paper!

It was a lot of fun collaborating with my coauthors @arttestt and others.

Of course, it wouldn't be possible without the ultralytics package maintainers like @onuralpszr and rest of the team.

[fatih c. akyon @fcakyon]

@b_nnett App Store doesn’t allow reverse engineered apps. How do you plan to publish there?

[fatih c. akyon @fcakyon]

Announcing SAHI v0.12 from a rooftop in Denver, where CVPR 2026 kicks off tomorrow 👌 One of our biggest releases yet:

• Up to 7x faster inference on large images, with batch inference and new postprocessing backends
• Open-vocabulary detection and segmentation: @ultralytics YOLOE, YOLO-World, YOLO26, GroundingDINO and more
• Torch-free core, so you install only what you need
• Docs now in English and 中文

Most of the heavy lifting on this one was done by @onuralpszr. Also big thanks to @muhammdrizwanmr plus everyone else who contributed.

At CVPR this week? Let's meet 😎

pip install -U sahi

[fatih c. akyon @fcakyon]

😮

[Andrej Karpathy @karpathy]

Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.

[fatih c. akyon @fcakyon]

Claude Code'un geliştiricisi Ankara meetup'ımızda ekranını paylaştığında, eş zamanlı çalıştırdığı session sayısı herkesi şaşırttı. Cevap ekteki videoda 😅

Aynı meetup'ta her gün @Ultralytics'te kullandığım plugin marketplace üzerine bir konuşma yaptım: claude-codex-settings. O günden bu yana GitHub'da 700 yıldızı geçti ve yalnızca son 14 günde 3000'den fazla indirme aldı!

İçinde neler var:
- daha iyi context sıkıştırma ve session takibi için gizli configler
- @ultralytics'teki tam zamanlı YOLO işimde her gün kullandığım hook ve skill'ler
- Hetzner, Dokploy, GitHub, Overleaf ve daha fazlası için hazır skill'ler
- Claude Code, Cursor, Codex CLI ve Gemini CLI ile uyumlu

Klonlamak, incelemek isteyen için: https://t.co/47CCZ1aort

[fatih c. akyon @fcakyon]

When the creator of @Claude Code (@bcherny) shared his screen at our Ankara meetup, the number of parallel sessions he had running surprised everyone. The answer is in the attached video 😂

I gave a talk at the same meetup on claude-codex-settings, a plugin marketplace I built and use every day at @Ultralytics. Since then it has crossed 700 GitHub stars with 3,000+ downloads in the last 14 days alone.

What is inside:
- hidden configs for better context compaction and session monitoring
- the exact hooks and skills I rely on daily for full-time YOLO work at @Ultralytics
- ready-made skills for @Hetzner_Online, @getdokploy, GitHub, @overleaf, and more
- works with Claude Code, @cursor_ai, Codex CLI, and @GeminiApp CLI out of the box

Clone, inspect, fork what you like: https://t.co/47CCZ1aWh1

Huge thanks to @alpsencerozturk for organizing the meetup 🙏

[fatih c. akyon @fcakyon]

17 Aralik'ta Claude Code Ankara bulusmasindayiz 💯

@AnthropicAI ve @BilkentAI Society ortakliginda düzenlenen etkinlikte Claude Code bas muhendisi @bcherny'den hemen once konusacagim! @Ultralytics'te ve doktoramda kullandigim Claude Code ayarlarimdan ve gelistirdigim Claude Code pluginlerimden bahsedeceğim 👌

Katilim sayiyla sinirli o yuzden ilk yorumdaki linkten kayit olmayı unutmayın 👇

---

We're hosting a Claude Code Ankara meetup with @AnthropicAI, featuring Head of Claude Code Boris Cherny online! I'll discuss my @Claude Code settings used in Ultralytics and my PhD. Send me any feature requests or questions for Boris to pass on 👌

[Ultralytics @ultralytics]

New tutorial | Small object detection with SAHI and Ultralytics YOLO26 🔍

Learn how to use slicing-aided inference with YOLO26 to improve the detection of small and distant objects in large images.

Watch here ➡️ https://t.co/Fo8Q9fm9G5

#Ultralytics #YOLO26 #SAHI

[Ultralytics @ultralytics]

Run SAHI tiled inference with Ultralytics YOLO26! 🧩

Split large images into smaller tiles to improve small object detection accuracy, then merge the results for precise inference in aerial imagery, surveillance, and dense-scene analysis.

Learn more ➡️ https://t.co/iDM6c3255k

#Ultralytics #VisionAI #Sahitiledinference

[fatih c. akyon @fcakyon]

anthropic removed thinking blocks from claude code vscode extension on 16th april.

if you want to see still see the thinking blocks (i found them very important), this trick works: https://t.co/S1T4ZE92bX

[Muhammad Rizwan Munawar @muhammdrizwanmr]

Detect small objects using @Ultralytics YOLO26 + SAHI 😍

A perfect combo when speed isn’t the priority, but accuracy is everything, especially when detecting tiny or distant objects in your images.

learn how it works 👇

#AI #objectdetection #trafficanalysis

[fatih c. akyon @fcakyon]

See you in the @ultralytics booth, in Denver, this June 💯

[fatih c. akyon @fcakyon]

🔥 Presenting SenBen at CVPR Workshops 2026: https://t.co/hWMlhMBLSx

A 241M model just beat every frontier VLM we tested except Gemini, on grounded scene graph metrics for content moderation.

Most moderation systems return a binary safe/unsafe label without telling you what was detected, who is in the frame, or where it occurs. SenBen turns moderation into a structured prediction task: 13,999 movie frames annotated as Visual Genome-style scene graphs, 16 sensitivity tags, and affective attributes like pain, fear, aggression, and distress. The student is distilled from Gemini 3 Pro using a Vocabulary-Aware Recall loss and a decoupled tag head.

Full open release under MIT: dataset, training code, the VAR loss, SenBen-Score, all 18 baselines. Almost every safety pipeline at this scale is closed, which makes it impossible to compare or build on. We wanted to do the opposite.

#ExplainableAI #AISafety #CVPR2026 #VLM

https://t.co/JD810GvtOX

[fatih c. akyon @fcakyon]

I have just open-sourced a literature summary on ML-based content moderation and multimodal content rating! It includes various sources for audio, text, video modalities, links to datasets and related Python tools.

Github: https://t.co/memn97CK5F

Arxiv: https://t.co/WK6e126MP7

[fatih c. akyon @fcakyon]

https://t.co/3AeJsqYF9N