Home

Awesome

DLI-Net

Pytorch implementation of "DLI-Net: Dual Local Interaction Network for Fine-Grained Sketch-Based Image Retrieval". DLI-Net is the improved version of "DLA-Net for FG-SBIR: Dynamic Local Aligned Network for Fine-Grained Sketch-Based Image Retrieval". So you can also refer this code to implement DLA-Net.

Environment

Structure

image

Datasets

There are six datasets used in our paper. To train our DLI-Net, you need change the parameters of *.sh (e.g. ph_train_root and ph_train_txt). It is important to note that the folder structures of datasets need to be consistent with ours, otherwise you need to modify the code accordingly. Our folder structures are shown below.

Training

First modify the parameters in *.sh files, then use bash *.sh to train model.