Datasets:

kalyan1729
/

sentineledgedataset

license: cc-by-4.0
task_categories:
  - object-detection
  - video-classification
tags:
  - traffic
  - vehicle-detection
  - vehicle-tracking
  - cctv
  - bengaluru
  - urban-traffic
  - computer-vision
  - autonomous-driving
  - multi-object-tracking
  - indian-traffic
pretty_name: SentinelEdge Traffic Dataset
size_categories:
  - 100B<n<1T

SentinelEdge Traffic Dataset

A large-scale urban traffic dataset combining Bengaluru CCTV footage with the UA-DETRAC benchmark, assembled for the SentinelEdge traffic management and vehicle detection system. Contains ~196,000 annotated images across four sub-datasets covering Indian urban roads and international highway scenarios.

Sub-datasets

1. BMD-45-Train

35,792 images from 45 Bengaluru CCTV locations
373,132 bounding box annotations in COCO JSON format
Covers mixed urban traffic including autos, two-wheelers, heavy vehicles

2. UVH-26-Train

17,387 images from 26 urban Bengaluru camera locations
COCO JSON annotations (single-track and multi-view variants)
High vehicle density, diverse lighting conditions

3. IITM-HeTra v2

2,836 images from IIT Madras Heterogeneous Traffic dataset
Pascal VOC XML annotations
Train/val/test splits provided via trainval.txt and test.txt

4. UA-DETRAC

140,131 images from 100 traffic sequences recorded in Beijing and Tianjin
XML annotations with per-vehicle attributes (type, occlusion, truncation)
Includes MOT evaluation toolkit

Dataset Structure

├── BMD-45-Train/
│   ├── images_000/                  # 35,792 images (.png)
│   └── _annotations.coco.json      # COCO format annotations
│
├── UVH-26-Train/
│   ├── data/
│   │   ├── 000/                     # image sequences
│   │   ├── 001/
│   │   └── 002/
│   ├── UVH-26-ST-Train.json         # single-track annotations
│   └── UVH-26-MV-Train.json         # multi-view annotations
│
├── IITM-HeTra_v2/
│   ├── Dataset-1/
│   │   ├── images/
│   │   ├── images_test/
│   │   ├── xmls/                    # Pascal VOC train annotations
│   │   ├── xmls_test/               # Pascal VOC test annotations
│   │   ├── trainval.txt
│   │   └── test.txt
│   └── Dataset-2/
│       ├── images/
│       ├── images_test/
│       ├── xmls/
│       ├── xmls_test/
│       ├── trainval.txt
│       └── test.txt
│
└── DETRAC/
    ├── DETRAC-Images/               # 140,131 images across 100 sequences
    ├── DETRAC-Train-Annotations-XML/
    ├── DETRAC-Test-Annotations-XML/
    └── DETRAC-MOT-toolkit/          # MATLAB evaluation toolkit

Vehicle Classes

BMD-45, UVH-26, IITM-HeTra (Indian traffic):

Class	Description
Two-wheeler	Motorcycles, scooters
Three-wheeler	Auto-rickshaws
Hatchback	Small passenger cars
Sedan	Mid-size passenger cars
SUV	Sport utility vehicles
MUV	Multi-utility vehicles
Van	Vans
LCV	Light commercial vehicles
Mini-bus	Small buses
Bus	Full-size buses
Tempo-traveller	Tempo travellers
Truck	Heavy trucks
Bicycle	Bicycles
Others	Miscellaneous (UVH-26 only)

UA-DETRAC (Chinese highway traffic): Car, Bus, Van, Others

Statistics

Sub-dataset	Images	Annotations	Format	Region
BMD-45-Train	35,792	373,132	COCO JSON	Bengaluru, India
UVH-26-Train	17,387	—	COCO JSON	Bengaluru, India
IITM-HeTra v2	2,836	—	Pascal VOC XML	Chennai, India
UA-DETRAC	140,131	—	XML + MOT	Beijing/Tianjin, China
Total	~196,000	370,000+

Download

Full dataset:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="kalyan1729/sentineledgedataset",
    repo_type="dataset",
    local_dir="./sentineledgedataset"
)

Single sub-dataset:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="kalyan1729/sentineledgedataset",
    repo_type="dataset",
    local_dir="./detrac",
    allow_patterns="DETRAC/**"
)

Citation

If you use this dataset, please cite the original sources:

UA-DETRAC: Wen et al., "UA-DETRAC: A New Benchmark and Protocol for Multi-Object Detection and Tracking", 2020.
IITM-HeTra: IIT Madras Heterogeneous Traffic Dataset v2.

License

Creative Commons Attribution 4.0 International (CC BY 4.0)

Dataset compiled for the SentinelEdge traffic intelligence project.