Unsupervised Deep Imputed Hashing(UDIH) for partial cross-modal retrieval (IJCNN2020)- Tensorflow

Prerequisites

Linux or Windows

Python 3

Tensorflow

Numpy

Getting started

Clone this repo:

git clone https://github.com/AkChen/UDIH
cd UDIH

Downlowd the dataset (MIRFlickr & NUS-WIDE)

https://pan.baidu.com/s/1A9ZLU8l-PKJ0xN8kLWkAFQ  pwd:j5gs 

Put the .mat file under <UDIH_DIR>/data/

Generate Partial Data

cd PCMH-mir
python  generate_partial_data_mir.py

Once the partial data is generated, you just need focus on Imputation and Hashing learning.

Imputation

Set P_i (default: 0.02) and P_t (default: 0.01) in PCMH_2Path_Imputation_MIR.py

python PCMH_2Path_Imputation_MIR.py

Hashing Learning

Set the length of hash code (default: 16) BIT

python PCMH_Hashing_Leaning_MIR_16.py

The results will be recorded in <UDIH_DIR>/record