Top data sets for autonomous driving for first-class projects


Recently, more and more organizations and exploration foundations have made their autonomous driving records accessible to people. However, the best data sets on autonomous driving are not always simple and easy to find. That’s why we’ve presented you here at Analytics Insight with the best data sets on autonomous driving for your first-class projects.

What is a driving data set?

Driving data sets are those that essentially consist of data that is recorded with multiple sensors such as cameras, LIDARs, radars and GPS in a variety of traffic scenarios at different times of the day and different weather conditions and locations.

Now that we understand the basic definition, let’s take a look at the best autonomous driving datasets for your world-class projects.

1. A2D2 data set for autonomous driving

The Audi Autonomous Driving Dataset (A2D2) supplied by Audi was provided to help new businesses and academic researchers move forward in autonomous driving. The data set includes more than 41,000 labeled with 38 elements. A total of approx. 2.3 TB, A2D2 is separated by the annotation type (e.g. semantic subdivision, 3D bounding box). Regardless of the labeled data, A2D2 delivers unlabeled sensor data (~ 390,000 frames) for sequences with few loops.

2. Open ApolloScape data set for autonomous driving

As part of the Apollo Autonomous Driving Project, ApolloScape is an on-going exploration project that aims to encourage advances in all areas of autonomous driving, from perception to route to control. Via their website, customers can examine a selection of recreational devices and over 100,000 street view frames, 80,000 lidar point clouds and 1,000 km of trajectories for city traffic.

3. Argoverse record

Argoverse is made up of two sets of data designed to support autonomous vehicle ML companies, such as 3D tracking and motion prediction. The dataset was collected from a fleet of autonomous vehicles in Pittsburgh and Miami and includes 3D tracking annotations for 113 scenes and more than 324,000 unique vehicle trajectories for motion prediction. Very different from most other open source autonomous driving datasets, Argoverse is the modern AV dataset that provides forward-looking stereo images.

4. Berkeley DeepDrive data set

The DeepDrive data set, also called BDD 100K, offers customers access to 100,000 annotated videos and 10 tasks for evaluating image recognition algorithms for autonomous driving. The data set comprises over 1000 driving hours with more than 100 million frames as well as data on geographical, ecological and climatic variability.

5. CityScapes dataset

CityScapes is an enormous data set that focuses on the semantic understanding of urban street scenes in 50 German metropolitan areas. It highlights semantic, event-related, and thick pixel comments for 30 classes that are grouped into 8 classifications. The entire data set contains 5,000 explained images with fine explanations and 20,000 additional annotated images with rough comments.

6. Comma2k19 record

This dataset includes 33 hours of driving time recorded on Highway 280 in California. Each minute-long scene was recorded on a 20 km stretch of parkway between San Jose and San Francisco. The information was collected using comma EONs, which include a street-facing camera, phone GPS, thermometer, and a 9-axis IMU.

7. Google Landmarks Record

The Landmarks dataset, distributed by Google in 2018, is divided into two sets of images to assess the recognition and restoration of man-made and natural landmarks. The first data set contains more than 2 million images representing 30,000 special milestones from around the world. In 2019, Google distributed Landmarks-v2, a significantly larger data set with 5 million images and 200,000 sights.

8. KITTI Vision Benchmark Suite

The KITTI dataset was first used in 2012 by Geiger et al. and provided with the aim of advancing autonomous driving research with a set of benchmarks for real-world vision. As one of the very first data sets on autonomous driving, KITTI shines with more than 4000 scientific references and even more.

KITTI offers 2D, 3D and elevated perspective article recognition records, 2D article and multi-object tracking and segmentation records, road / lane evaluation records, both pixel and semantic records at the instance level, as well as raw records.

9. LeddarTech PixSet data set

Leddar PixSet, which will be launched in 2021, is a new publicly available data set for research on autonomous driving that contains data from a full suite of AV sensors (cameras, LiDARs, radar, IMU) and full waveform information from the Leddar Pixel, a 3D. , contains Strong State Streak LiDAR sensor. The data set contains 29,000 edges in 97 groupings with more than 1.3 million annotated 3D boxes.

10. Level 5 Open Data

The Level5 dataset is distributed by the famous ridesharing application Lyft and is another exceptional hotspot for autonomous driving data. It contains more than 55,000 human-labeled 3D comments on outlines, a surface guide, and a basic HD spatial semantic map captured by 7 cameras and up to 3 LiDAR sensors that can be used to contextualize the data.

Share this article

Do the sharing thing

Source link

Leave A Reply

Your email address will not be published.