2022 Challenge Tracks

Detailed participant instructions can be accessed here.

Participants can compete in one or more of the following four challenges:

Challenge Track 1: 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 2: Tracked-Vehicle Retrieval by Natural Language Descriptions

Natural language (NL) description offers another useful way to specify vehicle track queries. In this challenge track, participating teams will perform vehicle retrieval given tracks of vehicles and the NL descriptions of the target vehicles that describe both static properties of the target (e.g. vehicle size, type, and color) and dynamic properties of the target (e.g. vehicle motion and its relations to other vehicles or the environment). The performance of the retrieval by NL task will be evaluated using standard metrics of retrieval tasks while considering ambiguity caused by similar vehicle types, colors, and motion types.

Challenge Track 3: Naturalistic Driving Action Recognition

Distracted driving is highly dangerous and is reported to kill about 8 people every day in the United States. Today, naturalistic driving studies and computer vision techniques provide the much needed solution to identify and eliminate distracting driving behavior on the road. However, lack of labels, poor data quality and resolution have created obstacles along the way for deriving insights from data pertaining to the driver in the real world. Naturalistic driving studies serve as an essential tool in studying driver behavior in real-time. They capture every action of the driver in the traffic environment such as those involving drowsiness or distracted behavior. In this challenge track, users will be presented with synthetic naturalistic data of the driver collected from multiple camera locations inside the vehicle. The objective is to classify the distracted behavior activities executed by the driver in a given time frame. The training dataset will consist of a diverse group of drivers with and without performing 18 different tasks (such as phone call, eating, and reaching back) that could potentially distract them from driving. The performance of this classification task will be evaluated in terms of the speed and accuracy of the model. Participating teams will have the option to use any one camera view for the classification of driver tasks.

Challenge Track 4: Multi-Class Product Counting & Recognition for Automated Retail Checkout

A growing application of AI and computer vision is in the retail industry. Of the various problems that can be addressed, this track focuses on accurate and automatic check-out in a retail store. The challenge stems from the real-world scenario of occlusion, movement, similarity in items being scanned, novel SKUs that are created seasonally, and the cost of misdetection and misclassification. Participating teams will count and identify products as they move along a Retail checkout conveyor belt. For example, given a conveyor belt snapshot/video teams will count and identify all products that follow pre-defined movements. Products may be occluded or very similar to each other. 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. Evaluation will be on a test set of objects not included in training for both for the closed and open-world scenarios. To maximize the practical value of the outcome from this track, both product recognition effectiveness and the program execution efficiency will contribute to the final score for each participating team. The team with the highest combined efficiency and effectiveness score will be declared the winner of this track.