Neural-Nets for age-prediction from features extracted from drawings with VGG19 architecture with weights pre-trained on ImageNet
Author: Pablo Caceres
Created: 04/01/19
Updated: 05/20/19
Repo containing images of human drawings from children and adults
Here we use the VGG-19 ConvNet architecture with weights trained on ImageNet to extract features from intermediate layers. Such features are used to build and train a two neural-nets: (1) linear prediction of age; (2) binary classification of age (adult vs child). We compare the performance of the neural nets with a set of standard machine learning classifiers.
- Tensorflow 1.13.1
- Python 3.6
- Python libraries: Numpy, os, Scipy, matplotlib, Sklearn, Pandas, seaborn, pathlib, rarfile
Two options:
- Locally: run jupyter notebook
- Cloud: run in google colab
Work in progress