FearNot!
Introduction:
Kunj Joshi and I teamed up for Northeastern ACM's HackHealth'26 hackathon where 3 sponsoring start-ups challenged us to build off of their respective, physical, mental, and educational health-centric products.
Kunj and I immediately gravitated towards SkipIt, the mental health startup, whose product is designed to detect user-specific traumatizing scenes from movies, tv shows, or other content, in advance to warn sensitive users and enable them to blur the sensitive components or skip the scene altogether. Their mission is incredibly noble and we were very excited by the AI application prospects that tie into their tech.
​
We won!
FearNot!:
FearNot! is intended to be a proof of concept as to how SkipIt may be able to further generalize and personalize the detection of sensitive material. No two people have the same sensitivities, and their sensitivity may not be easily categorized into an overarching theme such as sexual assault or abuse.
FearNot! is designed to allow users to provide their own unique triggers and feed an object detection pipeline to search for those triggers in advance. In theory, once a trigger is detected, SkipIt's pipeline would take over.
​
Collaborative project with Kunj Joshi for Northeastern ACM's HackHealth'26.​​​
Code
Huggingface Demo
Disclaimer:
The API call to an LLM and the YOLO-World model bog the Huggingface demo significantly. Refer to the video for a better demonstration opportunity.
​
The code requires an API key as well and may need tinkering. Reach out if you would like a hand with your own demo!