I was looking at the TensorFlow 2.0 Detection Zoo recently and found the SSD MobileNet V2 FPNLite 320x320
pre-trained model and was wondering what the FPN part in "FPNLite" means/stands for.
Asked
Active
Viewed 7,836 times
5

Mr. Ace
- 335
- 3
- 14
2 Answers
3
It stands for Feature Pyramid Network. Its a subnetwork which outputs feature maps of different resolutions. An explanation of FPN using detectron2 as an example is here: https://medium.com/@hirotoschwert/digging-into-detectron-2-part-2-dd6e8b0526e

katsu
- 604
- 6
- 7
-
Hello, could you take a look at this: https://stackoverflow.com/questions/64621656/can-we-show-underfitting-with-mobilenet – Oct 31 '20 at 12:28
-
3What does 320 x 320 means ? – Karunesh Palekar Nov 26 '20 at 14:24
-
1The resolution of the imput image – Florida Man Jan 29 '21 at 08:30
1
A Feature Pyramid Network, or FPN, is a feature extractor that takes a single-scale image of an arbitrary size as input, and outputs proportionally sized feature maps at multiple levels, in a fully convolutional fashion. This process is independent of the backbone convolutional architectures.

Deepak Raj
- 462
- 5
- 15