myEye2Wheeler : A Two-Wheeler Indian Driver Real-World Eye-Tracking Dataset
Bhaiya Vaibhaw Kumar, Deepti Rawat, Tanvi Kandalla, Aarnav Nagariya, Kavita Vemuri · 2025
Read myEye2Wheeler : A Two-Wheeler Indian Driver Real-World Eye-Tracking Dataset by Bhaiya Vaibhaw Kumar, Deepti Rawat, Tanvi Kandalla, Aarnav Nagariya, Kavita Vemuri (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-12723.txt.
Introduction
Five authors release myEye2Wheeler, a dataset of where motorcycle and scooter riders actually look while driving in Indian traffic. Nearly all gaze datasets come from car drivers on orderly roads, which misses the chaos of mixed traffic. The team's recordings capture the eye movements of two-wheeler riders as they weave through real streets, giving a direct view of how they scan for hazards and make choices.
The test of existing tools was blunt. A leading saliency model, TASED-Net, predicted rider attention far worse here than on a European car dataset. That gap, the authors argue, proves attention models must be fit to local traffic rather than imported. The dataset aims to support safer road design and lane planning for the riders who form the backbone of Indian transport.
Catalog record
| CREATOR | Bhaiya Vaibhaw Kumar, Deepti Rawat, Tanvi Kandalla, Aarnav Nagariya, Kavita Vemuri |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-12723
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
myEye2Wheeler : A Two-Wheeler Indian Driver Real-World Eye-Tracking Dataset Bhaiya Vaibhaw Kumar, Deepti Rawat, Tanvi Kandalla, Aarnav Nagariya, Kavita Vemuri Abstract This paper presents the myEye2Wheeler dataset, a unique resource of real-world gaze behaviour of two-wheeler drivers navigating complex Indian traffic. Most datasets are from four-wheeler drivers on well-planned roads and homogeneous traffic. Our dataset offers a critical lens into the unique visual attention patterns and insights into the decision-making of Indian two-wheeler drivers. The analysis demonstrates that existing saliency models, like TASED-Net, perform less effectively on the myEye-2Wheeler dataset compared to when applied on the European 4-wheeler eye tracking datasets (DR(Eye)VE), highlighting the need for models specifically tailored to the traffic conditions. By introducing the dataset, we not only fill a significant gap in two-wheeler driver behaviour research in India but also emphasise the critical need for developing context-specific saliency models. The larger aim is to improve road safety for two-wheeler users and lane-planning to support a cost-effective mode of transport. I INTRODUCTION Traffic safety is a global concern, with road accidents causing significant human and economic losses annually [ 1 ] . Two-wheeler drivers in countries like India are particularly vulnerable, navigating complex traffic environments [ 2 ] . Driver assistance technologies have improved safety, relying on understanding driver gaze behaviour [ 3 ] . However, research on two-wheelers, especially in chaotic traffic like India’s, is lacking [ 4 , 5 ] . The drivers in India face challenges [ 4 ] like dense traffic and unpredictable interactions like pedestrians and animals entering the lanes, sudden lane changes by other vehicles without indicators, wrong-side overtakes, vehicles of different engine capacity, un-synchronised traffic lights, potholes, etc. The physics and mechanics of driving and the driver-external world interaction are different from a 4-wheeler, in particular, the higher risk to life in an accident. Given the complexity, We advocate for specialised datasets to develop tailored Advanced Driver-Assistance Systems (ADAS). Existing datasets like DR(eye)VE and Look Both Ways [ 6 , 7 ] cater to four-wheeler drivers and thus fail to capture situations faced by Indian two-wheeler drivers and the decisions made by them, such as choosing to move through narrow gaps between other vehicles, frequently changing lanes, overtaking from the wrong side, all while also being aware of the aforementioned unpredictability of Indian roads. The myEye2Wheeler dataset addresses this gap. I-A Related Works I-A 1 Open Source Driver Gaze Datasets While many studies rely on simulators for data collection [ 8 , 9 ] , advancements in wearable eye trackers and vision processing algorithms have enabled real-world eye-tracking. This technology has diverse applications, such as investigating distractions from mobile usage [ 10 ] , the effects of stress on driving behaviour [ 11 ] , and assessing the efficacy of driver assistance systems [ 12 ] . Most driver eye-tracking research focuses on four-wheelers, with limited attention to two-wheelers [ 13 , 14 , 15 ] , outside Western countries. Developing robust datasets is crucial for refining prediction algorithms derived from ground-truth gaze estimation. This section reviews notable datasets to emphasise the unique contributions of the myEye2Wheeler dataset. I-A 2 Gaze Zone Predictor datasets The DG-UNICAMP dataset, by Ribeiro et al. [ 16 ] , integrates RGB, IR, and depth cameras, but its stationary vehicle setup limits capturing dynamic driving scenarios. The Driver Gaze in the Wild (DGW) Dataset, introduced by Ghosh et al. [ 17 ] , improves representation with a large subject pool but retains the limitations of a stationary setup. The Driver Monitoring Dataset (DMD) [ 18 ] focuses on driver behaviour and gaze zone classification of interior components of the car and, similar to previous datasets, does not include gaze dynamics to the external traffic or road. I-A 3 Traffic Driver Gaze Datasets The DADA-2000 dataset [ 19 ] is a benchmark for driver attention prediction in accident scenarios, comprising 2000 video sequences with detailed annotations covering diverse driving conditions and accident types. It provides fixation maps, saccade scan paths, accident categories, and spatial crash object locations, offering insights into the relationship between driver attention and accidents. The DR(eye)VE dataset [ 6 ] collected on European roads focuses on driver gaze research in varied lighting conditions. It includes 74 videos captured from a wearable eye tracker and a roof-mounted camera, providing synchronised gaze coordinates for each frame. The Look Both Ways (LBW) Dataset [ 7 ] offers gaze data from a wearable eye tracker and dashboard-mounted camera images. It provides insights into gaze patterns and analysis derived from the face images of drivers. The DGAZE dataset [ 20 ] captures traffic gaze data in Indian traffic conditions, featuring face and traffic images with gaze annotations, albeit collected in a lab setting mimicking 4-wheeler driving conditions. I-B Research Gap The majority of driver behavioural research concentrates on Western countries, where four-wheelers dominate, and two-wheelers are not the primary mode of transportation [ 21 ] . In these settings, traffic also tends to be more uniform compared to the conditions on Indian roads. Additionally, the driving approach in Western countries differs, as there is less traffic and a disciplined approach to lane switching or overtaking, leading to an almost automatic motor-memory, cognitive processing and attention allocation by drivers [ 22 , 23 ] . Research on real-world eye-tracking studies involving two-wheelers is nascent, primarily for vehicle flow-behavior analysis rather than assistive systems design or urban road planning. But a thorough analysis of the basis for strategic shifts in visual gaze and the decision-making factors, such as estimation of surround vehicle engine capacity, size of surround vehicles, the experience-based prediction of the other vehicle driver behaviour, traffic density, and drivers’ self-discipline, can extend the insights to better inter-vehicle interaction and road design. Our study aims to achieve four main goals: a) curating an eye-tracking real-world dataset for further research and analysis, b) extracting driver gaze behaviour, and c) developing a saliency model. This paper focuses on the initial step of curating a dataset to advance research in the field and support the accomplishment of the remaining goals. II DATA ACQUISITION Fig. 1: Flowchart representing the key steps in this study II-A Participants A total of 40 participants volunteered for the study (age range: 19-50 years; male = 34, female = 6). Participants used either their personal two-wheelers or were provided with a TVS Jupiter, a gearless scooter with a 109.7 cc single-cylinder air-cooled engine. All participants held valid two-wheeler driving licenses issued by the Regional Transport Office authority. Participants were categorised into two groups based on driving experience: ‘experienced drivers’ with over a decade of 2-wheeler driving and ‘novice drivers’ with 1 to 6 years of experience since obtaining a license. The dataset comprises 21 novice participants (age: 19-25 years; male = 18, female = 3) and 20 experienced participants (age: 27-50 years; male = 16, female = 4). Safety measures were rigorously enforced during the study. Participants were instructed to adhere to a speed limit of 60 km/h and follow all traffic rules. Both participants and the experimenter wore helmets throughout the ride. Participants were assured of anonymity and that their data would only be used for research. A nominal compensation of INR 250 was provided to acknowledge their