How could I implement Retina Net or feature pyramidal network for object detection by only using deep learning toolbox?

11 次查看（过去 30 天）
DaeYoung
DaeYoung 2020-8-10
回答： Raynier Suresh 2021-2-23
Since MATLAB 2020a, it now provides a new object network scheme such as SSD(single-shot detector), and YOLO v2
as a form of layer template.
Such templates make it easy to implement some typical detection networks, but they seem to be less flexible.
For example, when I want to implement my own detection network with multiple RPNs (from faster RCNN) that are connected to different conv layers,
MATLAB outputs erros because only one RPN should be included in detection network (faster RCNN).
The relatively new networks such as FPN, RetinaNet should have more flexibility to use multi-scale feature maps from conv layers.
How could I implement Retina Net or feature pyramidal network for object detection by only using deep learning toolbox?

请先登录，再回答此问题。

采纳的回答

Raynier Suresh
Raynier Suresh 2021-2-23
Hi, The deep learning toolbox allows you to create custom layers and custom training loops, using these features you can define your network architecture and then train it.
Refer the below links for more information
https://www.mathworks.com/help/releases/R2020a/deeplearning/ug/define-custom-deep-learning-layers.html
https://www.mathworks.com/help/releases/R2020a/deeplearning/ref/dlnetwork.html

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Object Detection 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by