2021 Challenge Tracks
Challenge Track 1: Multi-Class Multi-Movement Vehicle Counting Using IoT Devices
Participating teams will count four-wheel vehicles and freight trucks that follow pre-defined movements from multiple camera scenes. For example, teams will perform vehicle counting separately for left-turning, right-turning and through traffic at a given intersection approach. This helps traffic engineers understand the traffic demand and freight ratio on individual corridors, which can be used to design better intersection signal timing plans and apply other traffic congestion mitigation strategies when necessary. To maximize the practical value of the outcome from this track, both vehicle counting effectiveness and the program execution efficiency will contribute to the final score for each participating team. Additionally, mimicking the results of in-road hardware sensor-based counting systems, solutions to this problem are expected to run online in real-time. While any system can be used to generate solutions to the problem for general submissions, the final evaluation of the top methods will be executed using an IoT device. The team with the highest combined efficiency and effectiveness score will be declared the winner of this track.
Challenge Track 2: City-Scale Multi-Camera Vehicle Re-Identification
Participating teams will perform vehicle re-identification based on vehicle crops from multiple cameras placed at multiple intersections. This helps traffic engineers understand journey times along entire corridors. Like in the 4th edition of the challenge, the training set will be composed of both real-world data and synthetic data. The usage of synthetic data is encouraged as it can be simulated under various environments and can produce large training data sets. The team with the highest accuracy in detecting vehicles that appear in multiple cameras will be declared the winner of this track. In the event that multiple teams perform equally well in this track, the algorithm needing the least amount of manual supervision will be chosen as the winner.
Challenge Track 3: City-Scale Multi-Camera Vehicle Tracking
Participating teams will track vehicles across multiple cameras both at a single intersection and across multiple intersections spread out across a city. This helps traffic engineers understand journey times along entire corridors. The team with the highest accuracy in tracking vehicles that appear in multiple cameras will be declared the winner of this track. In the event that multiple teams perform equally well in this track, the algorithm needing the least amount of manual supervision will be chosen as the winner.
Challenge Track 4: Traffic Anomaly Detection
Participating teams will submit all anomalies detected in the test data, including car crashes, stalled vehicles based on video feeds from multiple cameras at intersections and along highways. The team with the highest average precision in anomaly detection in the submitted anomalies will be announced the winner of this track.
Challenge Track 5: Natural Language-Based Vehicle Retrieval
Natural language (NL) description offers another useful way to specify vehicle track queries. In this new Challenge Track, participating teams will perform vehicle retrieval given single-camera tracks and corresponding NL descriptions of the targets. The performance of the retrieval task will be evaluated using standard metrics of retrieval tasks while considering ambiguity caused by similar vehicle types, colors, and motion types.