To use this site, upload an image and press the Computer Vision button. The Yolo3-based model trained on the custom set and some of the 2 prepared classes will show up to you. This site is for quickly testing the Computer Vision on the Darknet on OpenCL engine done by Piotr Sowa. This site does not store any images on the server, only on your web browser session. All computation is done in memory only. That means you cannot share by URL effect of your computation try. However, you can right-click on the image computed and save it as your favorite web browser. In fact, processing of Full HD images takes only about 50ms on GPU. The rest is the overhead of the TCP/IP network speed. I wish you the nice experiment results!