Modifications to video processing pipelines, e.g., editlist generators, can cause the output videos to not match a user specification. This disclosure describes techniques for the automated verification of revisions to video editlist generators and/or video edit engines. A random set of edited videos and the corresponding edit specifications are selected (205). The number of videos in the set is determined based on an estimation of the likelihood of an error in the output video as a result of changes in the editlist generator. Editlists are generated for each selected video based on a control editlist generator and a modified editlist generator. Content fingerprinting techniques are utilized to compare videos generated based on the respective editlists and to determine whether the modified editlist generator produces expected results. A summary of comparison results (and optionally, video playback) is provided to enable both quantitative as well as visual evaluation.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Falk, Christian Ingemar; Starostina, Ekaterina; and Rozycki, Karol, "Evaluation and Verification of Video Editing Output Based on Content Fingerprinting", Technical Disclosure Commons, (January 29, 2021)