Active Terahertz Imaging Dataset for Concealed Object Detection

-Project Page & Dataset

---

Introduction

Concealed object detection in Terahertz imaging is an urgent need for public security and counter-terrorism. In this paper, we provide a public dataset for evaluating multi-object detection algorithms in active Terahertz imaging resolution 5 mm by 5 mm. To the best of our knowledge, this is the first public Terahertz imaging dataset prepared to evaluate object detection algorithms. Object detection on this dataset is much more difficult than on those standard public object detection datasets due to its inferior imaging quality. Facing the problem of imbalanced samples in object detection and hard training samples, we evaluate four popular detectors: YOLOv3, YOLOv4, FRCN-OHEM, and RetinaNet on this dataset. Experimental results indicate that the RetinaNet achieves the highest mAP. In addition, we demonstrate that hiding objects in different parts of the human body affect detection accuracy. The dataset will be available online.

Dataset

Object classes and quantity of the dataset.

Class	GA	KK	SS	MD	CK	WB	KC	CP	CL	LW	UN	In total
Item	Gun	Kitchen Knife	Scissors	Metal Dagger	Ceramic Knife	Water Bottle	Key Chain	Cell Phone	Cigarette Lighter	Leather Wallet	Unknown	--
Quantity	116	100	96	64	129	107	78	129	163	78	289	1349

Sample display for each category. The bottom row shows a zoomed-in view of the object.

Sample display for diversification.

Dataset structure:

/THZ_dataset_det_VOC
    /Annotations
        /D_N_F1_CK_F_LA_WB_F_S_back_0907140917.xml
        /D_N_F1_CK_F_LA_WB_F_S_front_0907140917.xml
        /D_N_F1_CL_V_LA_LW_V_RA_back_0907141138.xml
        /...
        /T_P_M6_MD_F_LL_CK_F_C_WB_F_RT_front_0906154134.xml
    /JPEGImages
        /D_N_F1_CK_F_LA_WB_F_S_back_0907140917.jpg
        /D_N_F1_CK_F_LA_WB_F_S_front_0907140917.jpg
        /D_N_F1_CL_V_LA_LW_V_RA_back_0907141138.jpg
        /...
        /T_P_M6_MD_F_LL_CK_F_C_WB_F_RT_front_0906154134.jpg

The filename format is as follows:

Item Quantity_Whether the item exists_Model Number_Item1 Name_Item1 Direction_Item1 Position_(Item2 Name_Item2 Direction_Item2 Position_Item3 Name_Item3 Direction_Item3 Position_) Front and Back ID_Timestamp.jpg

• Number of Items: S (Single Item), D (Double Item), T (Triple Item).
• Whether the item exists: P (item exists), N (item does not exist).
• Model numbers: M1 (male size 1), M2 (male size 2), M3 (male size 3), M4 (male size 4), M5 (male size 5), M6 (male size 6), F1 (female size 1) ), F2 (Female No. 2), F3 (Female No. 3), F4 (Female No. 4).
• Item name: GA (gun), KK (chopping knife), SS (scissors), MD (dagger), CK (ceramic fruit knife), WB (mineral water bottle), KC (keychain), CP (mobile phone), CL (lighter) ), LW (wallet).
• Item orientation: F (front), V (back), L (left), R (right).
• Item location: LA (left arm), RA (right arm), S (abdomen), C (chest), B (back), W (waist), N (hip), LT (left thigh), RT ( right thigh), LL (left calf), RL (right calf).
• Front and back identification: front (front image), back (reverse image).
• Timestamp: data collection time (month, day, hour, minute, second)

Download links:

• Google drive
• Baidu drive(Extraction code: x3od)
• NUAA drive

The division of the dataset can be found at:

• Baidu drive(Extraction code: 47td)

---

If you find this dataset useful in your research, please consider citing:

@misc{liang2021active,
      title={Active Terahertz Imaging Dataset for Concealed Object Detection}, 
      author={Dong Liang and Fei Xue and Ling Li},
      year={2021},
      eprint={2105.03677},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}