The problem VIgilAI solves

Facial recognition systems have been used by the police to solve crime for decades, however, the technology has a critical shortcoming. To be effective, the systems require a clear well-lit photo. Any blurry footage, faces turned at angles, or even minimal disguises like glasses or baseball caps significantly hamper its usefulness. Additionally, in many instances, the police only have a description of the suspect instead of any clear images. With only a description, the criminals could slip away… until now!

Secure.ai is a tool that scans a video library to find any individuals that match a given description. For example, the police get a witness account that a man in a brown jacket and tan pants just stole a backpack. No worries! The officer can submit that description into Secure.ai and view a dashboard of any relevant footage captured by CCTV.

Challenges we ran into

Challenges Encountered:

Model Selection: Determining the suitable model consumed significant time and energy. Multiple model trials ensued before opting for YOLOv5. We invested resources to curate a tailored dataset, essential for training our model effectively.

Dataset Training: Training the dataset presented challenges in achieving optimal accuracy. It took considerable time and effort to fine-tune parameters and methodologies to attain satisfactory results. Developing our own dataset was necessary to address specific project requirements and improve model performance.

OpenAI Clip Integration: Integrating OpenAI Clip posed complexities. Despite comprehensive documentation, implementation hurdles persisted. Diligent efforts were required to decode and resolve these challenges, significantly delaying the integration process.

Solutions Implemented:

YOLOv5 Model Selection: After extensive experimentation, YOLOv5 emerged as the preferred model. We dedicated resources to refine and fine-tune our model, leveraging a meticulously curated dataset to optimize performance.

OpenAI Clip Integration: Through collaborative problem-solving and persistent efforts, we successfully addressed the integration hurdles. Adhering closely to the OpenAI documentation, we dissected and resolved each obstacle encountered, ultimately achieving seamless integration.

Dataset Training: We invested significant time in experimenting with training methodologies to enhance accuracy. Through iterative adjustments and careful analysis, we refined our training process to achieve improved results. Developing a customized dataset enabled us to tailor training data to project needs, ultimately enhancing model performance and accuracy.