GaitSet

A flexible, effective and fast network for cross-view gait recognition. It consistent with the results in GaitSet: Regarding Gait as a Set for Cross-View Gait Recognition

We arrived Rank@1=95.0% on CASIA-B and Rank@1=87.1% on OU-MVLP.

Prerequisites

• Python 3.6
• PyTorch 0.4+
• GPU

Getting started

Installation

• (Not necessary) Install Anaconda3
• Install CUDA 9.0
• install cuDNN7.0
• Install PyTorch

Noted that our code is tested based on PyTorch 0.4

Dataset & Preparation

Download CASIA-B Dataset

ATTENTION

• Organize the directory as: your_dataset_path/resolutions/dataset_names/subject_ids/walking_conditions/views. E.g. gaitdata/64/CASIA-B/001/nm-01/000/. (We will update the code to be more compatible.)
• You should cut and align the raw silhouette by yourself. Our experiments use the align method in this paper.

Futhermore, you also can test our code on OU-MVLP Dataset. The number of channels and the training batchsize is slightly different for this dataset. For more detail, please refer to our paper.

Configuration

In config.py, you might want to change the following settings:

• WORK_PATH path to save/load checkpoints
• CUDA_VISIBLE_DEVICES indices of GPUs
• dataset_path (necessary) root path of the dataset (for the above example, it is "gaitdata")

Train

Train a model by

python train.py

Test & Evaluation

Use trained model to extract feature by

python test.py

• --iter iteration of the checkpoint to load
• --batch_size batch size of the parallel test

It will output Rank@1 of all three walking conditions. Note that the test is parallelizable. To conduct a faster evaluation, you could use --batch_size to change the batch size for test.