• 5 min read
GPT-4 Vision Prompt Injection
In this article, we explore what prompt injection is and the techniques people have been using to perform prompt injection attacks on GPT-4.
Blog
Piotr Skalski
ML Growth Engineer @ Roboflow | Owner @ github.com/SkalskiP/make-sense (2.4k stars) | Blogger @ skalskip.medium.com/ (4.5k followers)
Latest Posts by Piotr Skalski
• 6 min read
First Impressions with LLaVA-1.5
In this guide, we share our first impressions testing LLaVA-1.5.
• 11 min read
GPT-4 with Vision: Complete Guide and Evaluation
In this guide, we share findings experimenting with GPT-4 with Vision, released by OpenAI in September 2023.
• 8 min read
How to Train RTMDet on a Custom Dataset
Learn how to train a RTMDet computer vision model on a custom dataset.
• 7 min read
ChatGPT Code Interpreter for Computer Vision
In this article, we share the results of our experimentation with ChatGPT's code interpreter feature on various computer vision tasks.
• 7 min read
How to Train YOLO-NAS on a Custom Dataset
YOLO-NAS is the latest state-of-the-art real-time object detection model. Learn how to train YOLO-NAS on your custom data.
• 6 min read
Leveraging Embeddings and Clustering Techniques in Computer Vision
Explore the world of image embeddings in computer vision, as we dive into clustering, dataset assessment, and detecting image duplication. Discover dimensionality reduction techniques like t-SNE and UMAP. Use CLIP embeddings for analyzing image class distribution and identifying similar images.
• 5 min read
Zero-Shot Image Annotation with Grounding DINO and SAM - A Notebook Tutorial
In this comprehensive tutorial, discover how to speed up your image annotation process using Grounding DINO and Segment Anything Model. Learn how to convert object detection datasets into instance segmentation datasets, and use these models to automatically annotate your images.
• 6 min read
Grounding DINO : SOTA Zero-Shot Object Detection
Most object detection models are trained to identify a narrow predetermined collection of classes. Zero-shot detectors like Grounding DINO want to break this status quo by making it possible to detect new objects without re-training a model.
• 2 min read
Build Computer Vision Applications Faster with Supervision
Learn how Supervision, a new Python package with utilities for building computer vision apps, can help you work through your computer vision projects faster than ever.
• 4 min read
How to Code Non-Maximum Suppression (NMS) in Plain NumPy
SUMMARY
Double detection, where an object detection model fires multiple overlapping bounding boxes on the same object, is a common issue that breaks downstream analytics like counting and tracking. Non-Maximum Suppression (NMS) resolves this by computing Intersection over Union (IoU) across all predicted boxes and discarding lower-confidence duplicates
• 4 min read
Track and Count Objects Using YOLOv8
Counting moving objects is one of the most popular use cases in computer vision. It is used, among other things, in traffic analysis and as part of the automation of manufacturing processes. That is why understanding how to do it well is crucial for any CV engineer.
• 9 min read
How to Train YOLOv8 Object Detection on a Custom Dataset
In this article, we walk through how to train a YOLOv8 object detection model using a custom dataset.
• 11 min read
How to Train YOLOv7 Instance Segmentation on a Custom Dataset
In this article, we're going to walk through how to detect concrete cracks using instance segmentation.
• 4 min read
Track Football Players with Computer Vision
In this post, we use a YOLOv5 detection model and state-of-the-art tracker ByteTRACK to track football players.
• 4 min read
Use Docker to Deploy Computer Vision Models
SUMMARY
Deploying a computer vision model involves far more than training it: managing Python packages, CUDA versions, system-level dependencies, and hardware differences all create friction before a single inference runs in production. Docker addresses this by packaging an application and all its dependencies into a portable container image, which
Stay Connected