YOLO is another object detection algorithm. In this article, https://jonathan-hui.medium.com/real-time-object-detection-with-yolo-yolov2-28b1b93e2088, the description is a little bit more detailed. A CNN classifier uses a CNN to output a single score but there are no restrictions that it cannot output 25 values — instead of one channel output, you have 25 channels. Because I don’t know what is your background in the field, the YOLO will be the best starting point. But if you still have issues, you may need to spend more time on deep learning itself more. Unfortunately, that will be out of scope for this article. The first few lectures in Stanford’s deep learning class will then be a good start. Again, I don’t know your background, so hard to access your questions for a better response.