Repository files navigation

This is a TensorFlow implementation of our ECCV paper"A Modulation Module for Multi-task Learning with Applications in Image Retrieval".The released code include the training and testing code for 7 attributes. This code could support training and testing with any number of training attributes. The code can be easily changed for 20 and 40 number of attributes training. We have released the prepared training data and testing data for 7 attributes. The training data supports training with any number of attributes up to 40 attributes in the dataset. The testing data for other attributes will come soon.

The code is tested using Tensorflow 1.0.0 under Ubuntu 14.04 with Python 3.5. The code is based on the facenetimpelementation. You could refer it for the software requirements.

TheCelebA dataset has been used for training and testing. This training set consists of around ranmdomly selected 30,000 image pairs over 40 attributes after face detection. We have prepared the traininghere and testing datahere.

./experiments/train_mm.sh./experiments/test_mm.sh

You could find the trained model [here]