PDIWS: Thermal Imaging Dataset for Person Detection in Intrusion Warning Systems

The PDIWS dataset is available here or Google drive like here. Please refer to our paper for more details: https://ieeexplore.ieee.org/document/10208055 or https://arxiv.org/abs/2302.13293 (preprint).

Installation

Clone this repo or download source code

git clone https://github.com/thuan-researcher/Intruder-Thermal-Dataset.git

Install requirement packages

pip install requirements.txt

Download dataset and move it to the folder `Intruder-Thermal-Dataset`.

Data

Data description

Each image in the dataset is compounded by an object image and a background using the Poison image editing method (see ./PIE). The dataset consists of two subsets train and test:

train: 2,000 images under .JPG format, each image contains only one object.
test: 500 images under .JPG format, each image contains only one object.

Labels for those images are stored in files train.json and test.json. Each contains a list of annotation directories following the format:

{"image_id": i, "bbox": [x, y, w, h], "class": c}

where i is the image index (file name), bbox is the bounding box, class (0-4) is the type of position (0: creeping, 1: crawling, 2: stooping, 3: climbing, 4: other).