pyimagesearch.com / .pyimagesearch-com

Generate synthetic VQA data with BLIP and PaliGemma, convert it to Hugging Face datasets, and share it for easy access by the ML community.

Fine tune SmolVLM using DPO -- maximize human alignment with streamlined, effective preference optimization.

Run YOLO object detection models directly in the browser using ONNX, WebAssembly, and Next.js -- no server or GPU needed. Fast, private, and interactive.

Train YOLOv12 to detect and classify potholes by severity. Run it anywhere using Docker and deploy a visual interface with Gradio for image and video input.

Learn how to monitor a real-time people tracker with YOLOv12 and Centroid Tracker for efficient entry and exit counting.