Navigating the Challenges in Large-Scale 360° Street-Level Data Collection for Mappillary: A Case Study of Dar es Salaam, Tanzania
By Iddy Chazua
Since last year, the OMDTZ team partnered with HeiGIT gGmbH at Heidelberg University for a detailed 360-degree mapping project. During this collaboration, we successfully covered 40 wards, collecting more than 500,000 images across 1300 kilometres of roads. For OMDTZ, the primary goal was to explore how we could easily and cost-effectively update the existing OpenStreetMap platform. By uploading data into Mappillary, the platform’s AI automates data point generation, such as road signs, bus stations, amenities like shops, and other relevant information. Once these data points are generated, they are uploaded to OpenStreetMap, where the information can be continuously used for humanitarian efforts. The interest of HeiGIT gGmbH at Heidelberg University is in solid waste management, focusing on developing models for identifying street littering.
For mapping, OMDTZ always prioritises localized and cost-efficient methods. In this project, the team modified a tricycle by mounting the camera and charging system. The reason for using a tricycle was its ability to navigate in various terrains and narrow streets, making it ideal for areas with heavy traffic congestion and poor accessibility. Regardless of the success, the team encountered the following issues which could be used as learning:
- Narrow Street: The team opted to use tricycles due to their flexibility in navigating narrow roads and streets, which are often difficult for larger vehicles to access, However, in certain squatter areas, the streets are so narrow that even a tricycle or motorcycle struggles to pass through. To ensure that we still collected all the necessary data in these challenging locations, the team adapted by using handheld sticks to extend the reach of the camera and walking on foot, capturing as much information as possible.
- Safety: This was a primary concern, particularly in narrow quarter roads and low-populated areas where the team was more vulnerable. To mitigate these risks, the team consistently worked in groups and sought support from local community members or neighbourhood leaders, who provided invaluable insights into the area’s safety conditions. Through engagement with locals, the team gained a deeper understanding of the environment, which helped reduce the likelihood of encountering theft or other hazards.
- Weather Condition: This posed another challenge during the data collection process which could potentially damage the equipment or disrupt the workflow. To address this, the team took proactive measures such as ensuring all sensitive equipment, including cameras and mobile devices, were protected by weatherproof covers. Additionally, the team adapted their schedule to avoid extreme weather, opting for early mornings or late afternoons to avoid the midday heat.
- Tree Coverage: In some areas, dense tree coverage obstructed the camera’s ability to capture clear imagery of the roads and surrounding features. This was particularly problematic because the tree branches and canopies created shadows or blocked the line of sight and sometimes scratched the camera lens mounted on the top of the tricycle. To overcome this, the team adjusted the height and angle of the cameras to ensure better visibility of road signs, landmarks, and other essential features, in cases where tree coverage significantly impaired data collection, the team carefully planned alternative routes or used hand-held devices to capture imagery in areas where the camera on the tricycle couldn’t reach.
- Camera Positioning and Stability: The team faced the difficulty of capturing blurry images while manoeuvring the tricycle during the mapping exercise. Vibrations from the road surface, abrupt movements, and speed changes were the main causes of this problem. To address this issue the team decided to keep the mapping process moving at a steady or moderate pace. The camera mount’s orientation was adjusted and stabilized to reduce motion blur. Despite these safeguards, external factors including uneven terrain, road bumps, and sudden stops continued to present sporadic challenges. However, the team’s flexible strategy improved the street-level camera’s overall effectiveness and image quality.
- Traffic congestion: Capturing images in high-traffic areas leads to capturing more images in the same area since the camera captures images at intervals of 5 seconds
- Public Awareness & Privacy Concerns: Some residents may feel uncomfortable being recorded, requiring community and government engagement.
Apart from the Field challenges, the team also faces other challenges in data uploading and processing which include
- Internet Speed: The files collected are often large, and uploading them takes a significant amount of time. Due to the high bandwidth consumption, uploading during working hours usually disrupts internet access and affects the productivity of people in the office. To mitigate this issue, the team usually uploads the data during the night or on weekends, when the network is less congested.
- Data Storage & Backups: Due to the large size of the collected data, significant storage space is required both for the cameras’ memory and for backing up the data.
We extend our sincere gratitude to all our partners who have contributed to the expansion of our coverage, especially HeiGIT gGmbH at Heidelberg University for co-funding this initiative and Mappillary for their generous donation of a GoPro Max 360 camera. This collaboration has demonstrated the power of cost-effective, localized mapping solutions in enhancing OpenStreetMap and supporting solid waste management data. As this methodology continues to prove efficient, we aim to scale up these efforts to more cities in Tanzania and beyond. We remain open to new partnerships to further expand our impact, ensuring that open mapping continues to drive meaningful solutions for urban development, disaster preparedness, and strengthened collaborations.
