VIDEO FINGERPRINTING AND MATCHING
TECXIPIO REVERSE VIDEO SEARCH API - FAQs
APPLICATIONS OF REVERSE VIDEO SEARCH
What are applications of reverse video searches?
VIDEO FINGERPRINTING AND VIDEO MATCHING
VIDEO UPLOADS AND SEARCH VOLUMES
1. What type of API is the TECXIPIO Reverse Video Search API?
The TECXIPIO Reverse Video Search API is a stateless web API (HTTP/1.1) with HMAC Token Authentication.
2. What is a digital fingerprint?
For computers to understand visual content, it needs to be "translated" into a unique binary code. Therefore, machine learning algorithms are used to identify distinctive features of images or keyframes that are transformed into a unique code – the video's fingerprint, which represents the visual content.
In contrast to previous methods of identifying visual content, such as watermarking, the video fingerprinting technology does not add to or alter the video content.
3. What is a Reverse Video Search API?
With the TECXIPIO Reverse Video Search API, you are able to build your own Reverse Video Search Engine. Easily implement video fingerprinting technology into your internal processes and systems to conduct visual searches. In contrast to textual searches, where the metadata and tags of video files are compared, it allows you to identify videos based on the actual visual content.
4. Do I receive help with the implementation or setting up of a reverse search engine?
Our team is highly experienced in setting up the entire process. We would be happy to help if you need further assistance; please contact us to discuss your project.
APPLICATIONS OF REVERSE VIDEO SEARCH
1. Why do I need visual search technology to find videos?
Due to increasing digitalization, the use and distribution of video content has increased dramatically in recent years. However, with the growing demand and increasing number of videos uploaded, downloaded and streamed per second, (video) search engines such as Google, YouTube, Bing or Vimeo must also keep up.
The original video search options based on text searches, such as titles, metadata and tags describing the content, no longer do justice to the existing mass of video files and deliver inaccurate and sometimes vast quantities of "wrong" unsuitable hits. The use of video recognition software, which increasingly replaces the manual addition of tags with automatic content recognition and tagging, can only detect the visual content of the moving images to a very limited extent. In the recent past, neural networks are increasingly used to automatically recognize what an image or video is all about.
And although results have become much more precise, manual verification of search results is still an elementary and time-consuming part of filtering out the actual hits from the long lists of search results.
In the manual or partially automatic creation of text tags, search results still return a high number of false positives, while a large number of correct hits are not listed due to missing or incorrect tags. A video backward search compares videos based on the visual content, thus completely eliminating extensive and time-consuming text searches (taking into account different spellings, covering all tagging and labeling options, etc.) depending on the area of application, and replacing a large part of the time-consuming manual verification of long inaccurate result lists. The Reverse Video Search API generates digital fingerprints of your media content, compares all videos in the search results list with the original files and identifies matches - including heavily modified videos and low-quality files.
2. What are the applications of reverse video searches?
There are many applications where a large number of videos need to be compared and identified – video fingerprinting technology automates and therefore highly accelerates the detection of matches, saving valuable time and resources.
Identification of pirated content/illegal uploads
Soon after new TV shows or movies hit theatres or, at the latest, when they become digitally available through, for example, VOD/OTT providers, illegal copies of the entertainment content are uploaded to streaming websites. To prevent further distribution and to take illegal copies down, rights holders need to constantly search for relevant websites. However, search results deliver a high share of false-positive results and a reliable verification process is necessary before further steps can be taken. Today, the complete verification step is still conducted with a high share of elaborate manual work, making it a time- and resource-consuming task and therefore an expensive effort that rights holders have to make to protect their investment. The Reverse Video Search API can undertake large parts of the process to drastically reduce time and costs.
Track popularity/measure re-uploads
Streaming websites such as YouTube or Vimeo are popular platforms to promote, for example, new movies or songs. Tracking social interactions, such as likes, views, or shares of trailers and music videos, is crucial to assess popularity levels, measure promotional success, and, oftentimes, monetize the entertainment content. However, re-uploads to non-official channels dilute these numbers dramatically. This is why content owners need to find and track such copies. Not only are identical copies uploaded, but the original content is modified. For example, only certain sequences are shown or the original video is integrated as super- or sub-frames, which makes the verification of search results even more time-consuming. The video fingerprinting and matching technology of the API enables content owners to verify large numbers of potential copies, including modified versions.
Spam video filtering/upload filters
The above-mentioned applications are anti-piracy and tracking methods employed after copies of content have already been distributed online. Spam and upload filters are deployed to prevent such activities in the first place. The API can be implemented in internal systems. When a video file is uploaded to your systems, the Reverse Video Search API automatically generates a digital fingerprint of the content and matches it against your reference video database. Identified matches can then be blocked before going online.
Management of large media archives
Video searches in large video databases often deliver incomplete results. This is for several reasons; identical files are saved with different IDs and IDs are missing or inconsistent, which happens especially when the ID system has to be changed or developed over time. Conducting video searches using metadata or tags will not deliver reliable results, and, in the worst case, will produce a very long list of results containing a high share of false-positives due to being prone to false, missing, or mislabeled mappings and classifications. A reverse video search is based on visual search technology, which means that the content of the videos itself is compared with each other, eliminating the reliance on the abovementioned error-prone methods for video identification.
VIDEO FINGERPRINTING AND MATCHING
1. How does the Reverse Video Search API work?
The Reverse Video Search API generates digital fingerprints of your original (reference) videos and saves them on our servers. Whenever you upload search videos, the API also generates fingerprints for the search video and compares them against your database. With each upload, you can decide whether you want to apply the standard or advanced search and matching configuration. For more details, please see "What are standard/advanced searches?"
2. What videos can be identified as matches?
The matching algorithm will search for duplicate or altered videos; which alteration will be identified as a match depends on the settings you choose when uploading the search videos or if you have applied custom settings before uploading your reference videos.
In general, the API is not sensitive to color.
3. What are the different video modifications (alterations) and what do they look like?
Video modifications or alterations include mirroring, stretching, or squeezing. In this video, the different types of alterations are shown.
4. What are standard searches?
The TECXIPIO Reverse Video Search API always allows you to decide whether you want to perform standard or advanced searches. With this, it offers the flexibility to select the most appropriate configuration for each search request to optimize search times and costs.
A standard video search identifies duplicates and videos that have been modified through rotation, scaling, and brightness as matches.
5. What are advanced searches?
The matching algorithm of advanced searches will also include videos that have been modified through occlusions, sub-frames, and super-frames next to the results delivered with standard video searches.
6. What are custom settings/configurations?
Next to the standard and advanced search options, the video fingerprinting and matching software can be configured to identify additional alterations of videos. These include mirroring, stretching, squeezing, and combinations of these (e.g., videos that have been mirrored and stretched or mirrored and squeezed).
If you want to include any or all of these transformation types, please contact us regarding the pricing so we can set the API configurations according to your needs.
Please note that to conduct video searches with custom settings (additional transformation types), the digital fingerprints of the videos that you upload to your reference database will be affected. This means that if you have already created a reference database (collection) to which you want to apply additional transformation types at a later time, the videos need to be re-uploaded with the respective settings.
For more information please also check the functional documentation of the API.
VIDEO UPLOADS AND SEARCH VOLUMES
1. What happens with the videos I upload?
If you upload videos to the Reverse Video Search API, the files are temporarily stored in the working memory for the calculation of the digital fingerprints and removed from the working memory afterward.
2. Is there a limit to the size or file types of the videos I can upload?
There is generally no limit to the file size you upload. Video files may have the container formats MP4, Matroska, WebM, or AVI. Codec H.264 is preferred.
3. Is there a limit to the number of videos in the reference database?
No, there is generally no limit to the number of videos you can add to your collections or reference database. We would nevertheless recommend that you contact us if you plan to conduct high-scale reverse video searches that exceed a credit volume of 100,000,000 or €10,000/month so that we can look into special rates for you.
4. Can I use the API for high-volume searches?
The Reverse Video Search API is specially designed to perform high-volume searches at reasonable costs. Therefore, there is generally no limit to the number of reverse video searches you can conduct per month. We would nevertheless recommend that you contact us if you plan to conduct high-scale reverse video searches that exceed a credit volume of 100,000,000 or €10,000/month so that we can look into special rates for you.