Python project about fully convolutional neural network

Description

In this work, a first step towards the direct detection of “good” features using a fully convolutional neural network (CNN) is presented. By designing a convolutional light field autoencoder and comparing the results from 2D and 3D convolutions, we show that

convolutions can pick up on additional information in the light field in higher dimensions.
We then build a CNN to detect SIFT features on 2D input images. We show that the model architecture can be adapted to work with higher-dimensional inputs.

In this work, we present a learning approach for feature detection. We show that our method, using a fully convolutional neural network, is feasible for two-dimensional images and generalizes to the higher dimensional light fields.

Visualization of the model pipeline for detection. The original image is fed into the CNN which produces a heatmap that is used to detect features and compare them to features detected by SIFT

Learning curve of the 2D classification on rendered images

Python code for merging two separate photos to create a single, multi-person photo

Python Project

Python code for Classification of photographic images based on perceived aesthetic quality