Excerpts from good essays

Using FPN in a basic Faster R-CNN system, our method achieves state-of-the-art single model results on the COCO detection benchmark without bells and whistles, surpassing all existing single-model entries including those from the COCO 2016 challenge winners.

Personal understanding ：（without bells and whistles） There's not so much fancy , It refers to all kinds of without deep neural network trick, Anyone can do without all kinds of fine tuning , Expand the training set , hard online, Learning rate , Adjustment of parameters , Can achieve better results . So the essence of the method is good , Is inevitable , It is not accidental skills that improve the performance of the model .

— From abstract 2016 Characteristic pyramid FPN

---

3.2. Graph Convolutional Network Recap

—（Recap, Restate key points ）, From abstract 2019 ML-GCN The use of titles

---

The Illustrated Transformer
– illustrated ,Blog Title name in , The illustration Transformer. Excerpt from ：The Illustrated Transformer