# What is dense prediction in Deep learning?

I am using TensorFlow's pre-trained model of Convolutional Neural Network.

https://github.com/tensorflow/models/blob/master/slim/nets/resnet_v2.py#L130

I found following sentence:

However, for dense prediction tasks we advise that one uses inputs with spatial dimensions that are multiples of 32 plus 1, e.g., [321, 321].

Does anyone know what is dense prediction in this literature?

• "In computer vision pixelwise dense prediction is the task of predicting a label for each pixel in the image." semanticscholar.org/paper/…
– DaL
May 24, 2017 at 23:47

## 2 Answers

In computer vision pixelwise dense prediction is the task of predicting a label for each pixel in the image https://arxiv.org/abs/1611.09288

Agreeing to Thomas, Dense prediction task is producing dense output for the given input. In computer vision, unlike image classfication, tasks such as semantic segmentation, instance segmentation, etc. are considered as dense prediction task because label for each pixel is predicted.