Frequently Asked Questions Trace™ Q: How does Trace™ work? A: Trace™ contains two tools: fingerprint generation and matching. There are two usages of the tools: you can license both tools, build a database of fingerprints for the copyrighted contents, generate and compare the fingerprint of the suspected video clip against that database. You can also just license the client fingerprint generation tool to generates the fingerprints and send them to Vercury's fingerprint matching website which will give the verdict of the clip. Q: What are the false-positive/false-negative percentages for Trace™? A: It depends on the length and the quality of the suspected videos. Trace™ is able to find the match even if the suspected video has 3 seconds of copyrighted content as long as the video is viewable by human beings. For longer clips (a few minutes), Trace™ can find the match even if it is distorted. Q: How big are the fingerprints? A: For Flash videos, the fingerprints are usually less than 1/1000th of the original videos. For higher quality videos, they are even smaller. Also, it depends on the granularity of the detection, i.e. if you only intend to detect the longer clips, the fingerprints can be smaller. Those parameters are adjustable. Q: How fast is the fingerprint generation process? A: For Flash videos, it is usually 10-20 times faster than the playing time on a 1.5G Hz PC. Q: How fast is the matching process? A: In general, matching one clip (3 minutes) against 10,000 copyrighted videos takes less than a second when running on one PC. Again, this is related to the length of the copyrighted videos and the granularity of the detection. Q: How many codecs does Trace™ support? A: It supports most of the commonly used codecs, e.g. MPEG2/MPEG4, WMV, Flash, Quicktime, RealNetwork. Realtime Video Conversion™ Q: How does Realtime Video Conversion™ work? A: It stores one copy of the highest quality video stream, parses that video to get some critical data and stores those data in the system. When user requests the highest quality video, it sends the stored copy to the end user, but when a user requests the video in different resolution, quality or frame rate, it generates the requested video stream on the fly. Q: What are the advantages of this tool over the standard SVC approach? A: The standard is not finalized yet, and it only supports H.264. Also users need a new player to play SVC videos. Our tool supports most of the codecs and there is no need to change the player. Q: Does the tool increase CPU load significantly? A: We have a patented optimization techniques to speed up the video stream generation. Also, a typical LRU caching mechanism would help the performance too. DupScrubber™ Q: How does DupScrubber™ work? A: DupScrubber™ contains two tools: signature generation and matching. You use the signature generation tool to generate signatures for all the videos you have. When a new video is uploaded, you compare the signature of that video against all the existing signatures to see if it is a duplicate. Q: What are the false-positive/false-negative percentages for DupScrubber™? A: It is quite accurate as long as the video is not highly distorted. Q: How big are the signatures? A: Less than 100 bytes for any video. Q: How fast is the signature generation process? A: For Flash videos, it is usually 5-10 times faster than the playing time on a 1.5G Hz PC. Q: How fast is the matching process? A: Less than a second to compare one video against one million of videos. Q: How many codecs does DupScrubber™ support? A: It supports most of the commonly used codecs, e.g. MPEG2/MPEG4, WMV, Flash, Quicktime, RealNetwork.