Adding structured data to your website's pages is very important. This is because the rich results will continue to be used instead of the old-fashioned blue links. In Google's announcement that the beta version of the Rich Results Testing Tool will end, they said: "The rich result is a Google search experience that goes beyond standard blue links. They utilize structured data and can include carousels, images, or other non-text elements in the last two. Over the years, we have developed rich result tests that help us test structured data and preview rich results. " Tool improvements are welcome, and while there are quite a few tools and plugins that can help you generate structured data, there are still big gaps.
Adding the correct structured data to a web page remains manual and requires manipulation of the code, making it prone to human error. In ghost mannequin effect this post, I'll show you some fairly advanced techniques to help you generate high quality structured data without manual input. advertisement Continue reading below It relies on connecting the latest advances in computer vision with concepts from completely different disciplines. This is our technical plan: Take advantage of the list of screenshots on the product page annotated by my team to show the product name, price, description, main image and more. Converts annotations to the format expected by the Google AutoML Object Detection Service. Follow the steps to upload an annotated image, train it, and evaluate a model that can predict the annotations in the new screenshot.
Run the Python prediction code on the new image. Use puppeteer to automatically take new screenshots and insert JavaScript code to transform your predictions into structured data. Let's review some important concepts that make this technique possible. First, let's review some basic concepts. Computer vision 101 The best way to learn and understand the power of computer vision is to perform simple examples. We've put together a simple Colab notebook that basically performs the steps outlined in this repository of Streamlit. Make a copy and select Runtime> Run All to run all cells. Finally, you need to get a dynamic ngrok URL. Copy and paste it into your browser, download the YOLO weights, then select Run App in the sidebar to see the demo web app below.