TRECVID 2018 guidelines

Activities in Extended Video

Task coordinator: ActEV NIST team


The ActEV evaluation seeks to evaluate robust automatic activity detection algorithms for a multi-camera streaming video environment. ActEV is an extension of the annual TRECVID Surveillance Event Detection (SED) evaluation by adding a large collection of multi-camera video data, both of simple and complex activities. ActEV will address activity detection for both forensic applications and for real-time alerting.

At a high level, activity detection algorithms can be classified as either: unconstrained and constrained. The first evaluation will focus on the unconstrained condition only.

Unconstrained: For the latency, unconstrained condition, a system processes the whole video in the full corpus prior to returning a list of detected activity instances. The primary application of the evaluation condition includes a forensic investigation/analysis tool.

Constrained: For the latency, constrained condition, a system detects the presence of the target activity instance within a predefined duration (e.g., 2 seconds) from the time the target activity begins in a video streaming environment. A system on this evaluation condition supports a real-time alert application.

What is Activity Detection in Videos?

By activity detection, we mean to detect visual events (people engaged in particular activities) in a large collection of streaming video data. The ActEV evaluation is being conducted to assess the robustness of automatic activity detection for a multi-camera streaming video environment.

Who: NIST invites all organizations, particularly universities and corporations, to submit their results using their technologies to the ActEV evaluation server.  The evaluation is open worldwide. Participation is free. NIST does not provide funds to participants.

How: To take part in the ActEV evaluation you need to create an account on the ActEV scoring server (on the task website). Using a valid registered account, you will be able to upload your JSON format results (see website for: Evaluation Plan) to the ActEV scoring server and participate in the ActEV evaluation. Each team is limited to few submissions (see website for: Evaluation Plan). Only results that are submitted during the evaluation period and follow the Evaluation Plan will be considered valid and posted on the leaderboard website.

Why: The primary driver of the evaluation is to support investigations of crime by automatic activity detection in streaming video. These videos are of interest to NIST's partner agencies that seek to employ automatic activity detection in investigating of crime. ActEV evaluation seeks to develop robust automatic activity detection for a multi-camera streaming video environment. Activities will be enriched by person and object detection. ActEV will address activity detection for both forensic applications and for real-time alerting.


For information about the data used in the ActEV evaluation, please see website for: Licensing and Participation Agreements, and Evaluation Plan.
For provided development and testing data please see this guidelines data page


The main scoring metrics will be based on temporal localization and spatio-temporal localization. We will use standard evaluation measures (e.g., probability of missed detection, rate of false alarm, precision, recall, average precision) and self-reported speed (see website for: Evaluation Plan)

NIST Role: NIST will be conducting the ActEV evaluation which replaces the annual TRECVID Surveillance Event Detection (SED) evaluation.