Weapon Detection

The weapon data sets that are provided in this section focus specifically on the development of intelligent video surveillance automatic systems.

An automatic weapon detection system can provide the early detection of potentially violent situations that is of paramount importance for citizens security. One way to prevent these situations is by detecting the presence of dangerous objects such as handguns and knives in surveillance videos. Deep Learning techniques based on Convolutional Neural Networks can be trained to detect this type of object.

The weapon detection task can be performed by different approaches of combining a region proposal technique with a classifier, or integrating both into one model. However, any deep learning model requires to learn a quality image dataset and an annotation according to the classification or detection tasks.

Weapon detection Open Data provides quality image datasets built for training Deep Learning models under the development of an automatic weapon detection system. Weapons datasets for image classification and object detection tasks are described and can be downloaded below. The public datasets are organized depending on the included objects in the dataset images and the target task. 

The weapon detection datasets are available in the Open Data repository.

1 Weapon detection based on image classification

The datasets included in this section have been designed for the classification task based on CNN deep learning models. After the training stage on these datasets, the classification models must distinguish between weapons and different common objects present in the background or handled similarly.

The provided datasets are annotated by folders structure, where the images of a class are stored in a folder with the class name.

1.1 Handgun vs common objects

The dataset contains 9261 images splitted into 102 different object classes. The handgun class has 200 images, whereas the rest of classes include diverse common objects such as, airplane, chair, cell, animals, and so on.

This dataset is designed in the related publication giving additional information about the image dataset and experiment results.

Olmos, R., Tabik, S., & Herrera, F. (2018) Automatic handgun detection alarm in videos using deep learning. Neurocomputing, 275, 66-72 doi.org/10.1016/j.neucom.2017.05.012

The test dataset used in the work for performance measurement has a total of 608 images of which 304 are images of pistols.

1.2 Knives  vs common objects

The dataset contains 10039 pictures from the Internet and grouped into 100 different object classes. Knife is the target class containing 635 images apart from other handled objects such as pen, smartphone, cigar, or common objects in the background such as car, barrel, different plants, and some animals.

This dataset is designed in the related publication giving additional information about the image dataset and experiment results.

Castillo, A., Tabik, S., Pérez, F., Olmos, R., Herrera, F. (2019) Brightness guided preprocessing for automatic cold steel weapon detection in surveillance videos using deep learning. Neurocomputing, 330, 151-161. doi.org/10.1016/j.neucom.2018.10.076

1.3 Weapons and similar handled object

The dataset includes weapons and small objects that are handled in a similar way. It contains six different classes such as pistol, knife, bill, purse, smartphone and card. The classification images are obtained from the detection images in which the object’s bounding box has been cut out. 

The txt file gives information about the experiment dataset partitions used in the article.

This dataset is designed in the related publication giving additional information about the image dataset and experiment results.

Pérez-Hernández, F., Tabik, S., Lamas, A., Olmos, R., Fujita, H., Herrera, F. (2020) Object Detection Binary Classifiers methodology based on deep learning to identify small objects handled similarly: Application in video surveillance. Knowledge-Based Systems, 194, 105590. doi.org/10.1016/j.knosys.2020.105590

2 Weapon detection based on object detection

The datasets included in this section have been designed for the object detection task based on Deep Learning architectures with a CNN backbone. The selected images contain weapons and objects but also consider an enriched context of different background objects as well as the way objects are handled. After the training stage on these datasets, the detection models must locate and distinguish between weapons and different common objects present in the background or handled similarly.

The datasets also attached the annotation files in Pascal VOC format with the region of the target objects in xml files.

2.1 Handgun detection

The dataset contains 3000 images of short guns with rich context in the background. The images selected from the internet contain one or more handguns in diverse situations including video surveillance contexts.

This dataset is designed in the related publication giving additional information about the image dataset and experiment results.

Olmos, R., Tabik, S., & Herrera, F. (2018) Automatic handgun detection alarm in videos using deep learning. Neurocomputing, 275, 66-72. doi.org/10.1016/j.neucom.2017.05.012

2.2 Knife detection

The dataset contains 2078 images where at least one knife appears. The selected images were download from Internet, and some frames were extracted from Youtube videos or surveillance videos. The dataset take into account:

  1. cold steel weapon of diverse types, shape, colors, size, and made of different materials 
  2. knives located at different distances near and far from the camera
  3. knives occluded partially by the hand
  4. objects that can be handled in the same way as knives
  5. images captured in indoor and outdoor scenarios

This dataset is designed in the related publication giving additional information about the image dataset and experiment results.

Publication date:

June 2020

Contact:

Alberto Castillo Llama

Scroll Up