- CJ OliveNetworks and CJ ENM teamed up to embed an AI model on the video analytics platform to improve productivity.

- The AI model for video screening that is trained using large amount of data automatically detects inappropriate content such as smoking and violent scenes.

- The models check the similarity, citations, and frequency of usage against the original video to prevent the risk of excessive original IP exposure.

- With this technical partnership, the two companies plan to further advance the video screening model and stabilize the operating environment.

CJ OliveNetworks provides CJ ENM's video analysis platform with an AI model for video screening and similarity analysis (“video screening model”) to help improve work productivity.

CJ OliveNetworks has launched the AI Movie Filter service with CGV that can swap the faces of people in movie posters and completed various AI projects in the media industry, such as providing AI voice cloning technology for closed captioning for tvN.

The CJ ENM video analysis platform applies two technologies, the screening and the similarity analysis of the CJ OliveNetworks video screening model to greatly enhance the work speed without compromising the quality of content.

To create a complete video screening model, CJ OliveNetworks focused on increasing precision by ▲ building 12 APIs including object detection, video screening, and similarity ▲ applying content-specific algorithms and ▲ training video data.

The videos, clips, and images were labeled to train the model for screening, and self-verification tests were conducted.

The embedded video screening model on the CJ ENM video analysis platform automatically identifies and flags inappropriate content such as smoking, drinking, violence and sex. The model can distinguish objects such as cigarettes and alcohol and recognize behaviors such as profanity, violence, and sex. 

The new feature was added to the existing model to pinpoint whether a motorcycle rider is wearing a helmet or whether a driver in a car is wearing a safety belt as well as recognizing license plates to help with high-quality content production. 

In addition, the similarity analysis technology is applied to the platform for quick identification of things needing correction by comparing videos in the editing stage. It is also possible to quickly compare the edited video with the original one to see if the reviewer’s opinion was properly reflected.

Even a short clip can be screened against the original to check for similarity, citation, and frequency of footage usage to prevent excessive original IP exposure.

Using the video screening model, production teams can improve productivity because AI can focus on simple and repetitive tasks while team members can focus on higher-value activities.

The two companies will work on enhancing the model and stabilizing the operating environment to further expand automation of content management through AI technology.

“We proved that AI technology can provide real value in the field of content screening,” said Seok-ju Hong, the technology strategy manager at CJ OliveNetworks. “We plan to develop a large-scale video management service with video archiving and reproduction capabilities based on AI technology and apply it to various industries.”