Research project on dangers and safety of AVs with motorcycles. Addresses research gap by examining object classification methods, identifying potential solutions or issues.
1
stars
0
forks
source link
As a developer, I want to identify relevant data sources, so that collection of the project can commence. #9
An argument was raised: “Surely there must be a research focus on all types of road users before AVs went into production.” The lack of data sources involving Motorcycles may fall into one of these, excluding research focus:
Prevalence of Motorcycles: In many areas, especially those where AVs are being tested, motorcycles may be less prevalent than other road users, resulting in fewer instances of them in collected data.
Data Collection Challenges: As motorcyclists often move at higher speeds and in a less predictable manner than cyclists or pedestrians, obtaining high-quality sensor data can be more challenging. Noting that capturing motorcyclists’ or cyclists’ faces falls under data protection issues.
Proprietary Data: Companies may have substantial amounts of motorcyclist data, but they may choose not to make this data public due to competitive advantages or regulatory restrictions.
Some additional data sources involve:
Public Datasets: Numerous public datasets available could be relevant. Here are a few suggestions:
Waymo Open Dataset: Waymo, an autonomous driving company, has released a high-quality multimodal sensor dataset for autonomous driving.
nuScenes: This is a large-scale public dataset for autonomous driving provided by Aptiv (formerly nuTonomy). It includes data from various sensors and has several scenes of urban traffic.
Berkeley DeepDrive (BDD): This dataset contains various driving videos with different weather conditions and times of the day.
Argoverse: This dataset by Argo AI includes data from vehicles’ sensor suite, which provides for LIDAR and cameras.
Cityscapes: A dataset that focuses on semantic understanding of urban street scenes, it can be used for tasks like semantic segmentation or object detection in the context of self-driving.
Data collection retrieval could use APIs, scraping and partnerships.
Storing any datasets or sensitive documents within GitHub could cause Privacy Concerns, depending on the dataset’s or resource’s Terms of Use.
An argument was raised: “Surely there must be a research focus on all types of road users before AVs went into production.” The lack of data sources involving Motorcycles may fall into one of these, excluding research focus:
Some additional data sources involve: Public Datasets: Numerous public datasets available could be relevant. Here are a few suggestions:
Data collection retrieval could use APIs, scraping and partnerships.
Storing any datasets or sensitive documents within GitHub could cause Privacy Concerns, depending on the dataset’s or resource’s Terms of Use.