Screph: From Image to Working Program — Computer Vision

In this video, I introduce Screph—a desktop application that brings together image processing, computer vision, an AI agent, and code generation. Using three images of parts as an example, I walk through the process from defining the task to getting results: we select a processing region, try contour detection, spot unwanted objects, and refine the parameters with the agent’s help. We then save the tested configuration and use it in a program that processes images and writes the results to a table. Also covered in the video: • How Screph relates to graphical user interface automation. • Lighting correction and separating an object from the background. • Help from the agent with choosing processing methods and preparing the program. • Image annotation, correcting bounding boxes, and creating a dataset to train your own model. The core idea behind Screph is to start with your own task and image, test a solution, and then move on to repeatable processing. 00:00 — The task and the story behind Screph 01:07 — GUI automation and working with images 07:22 — Screph Code: the program and counting results 08:27 — Image annotation and dataset preparation Project website and documentation: screph.com Screph is under development. The images of parts were created specifically for this tutorial. The agent’s responses were generated in advance and are replayed in the recording; image processing and the prepared program run during the demonstration. Model training is not performed in this video. Photo used for annotation: “africa0703-0906a” — Dave Proffer, CC BY 2.0. A reduced-size copy is used, with bounding boxes and labels overlaid on the image. #Screph #ComputerVision #Automation