Doculist, currently in its pre-launch phase, is developing an innovative service designed to simplify the documentation of personal belongings for homeowners and renters. This service is particularly crucial for creating necessary documentation for insurance claims. Featuring intuitive mobile applications for both iOS and Android, Doculist utilizes interactive live camera scans of a user's space. As the scan progresses, the app detects objects and automatically populates as much information as possible about each item.

For any details not automatically captured, users can easily complete the remaining information. This can be done by either directly inputting purchase data or by syncing and matching the items to their online bank records. This seamless integration not only enhances the user experience but also ensures comprehensive and accurate documentation, vital for insurance purposes.

‍

Problem Statement

Existing asset management systems often face a significant challenge: the cumbersome and error-prone process of manually entering asset data, leading to low user adoption. This inefficiency leaves many homeowners with undocumented assets, resulting in the loss of billions of dollars in potential reimbursements due to inadequate proof of ownership.

Our goal is to alleviate this burden by transforming asset entry into a more engaging and effortless experience. We aim to gamify the process, making it both enjoyable and seamless for users. This innovative approach will not only simplify the documentation of assets but also ensure homeowners are well-positioned to maximize their potential reimbursement claims.

‍

‍

Solution

We have chosen to employ a dual-technology approach for our project. The primary method involves using smartphone cameras to conduct a real-time scan of a room. During this scan, the camera captures frames that are processed by an onboard Computer Vision model. This model, trained to recognize common household items, performs an initial room scan to identify distinct objects and categorize them (e.g., Television, Sofa, Painting). It also extracts observable attributes from the images, such as color, size, brand logos, serial numbers, and their location within the home.

These images are then uploaded to a more sophisticated AI system in the cloud. This cloud-based AI conducts a detailed analysis of the images and their extracted features to locate corresponding items in our extensive database of consumer products. The system suggests close matches for each detected object, offering users the option to confirm these matches or seek alternatives. Once a match is accepted, the system automatically retrieves and fills in the product's metadata, simplifying the asset documentation process.

Additionally, the system is designed to align each cataloged item with the user's bank records. This integration aims to import purchase details and locate the relevant receipt, which is crucial for insurance claims. Users are prompted to manually input any missing information to complete the asset documentation. This comprehensive approach ensures a seamless and efficient experience for users, significantly streamlining the asset documentation process.

‍

‍

Challenges

Our findings indicated that deploying local computer vision models necessitated the inclusion of substantial model files within application installations, coupled with the need for more powerful hardware. Also. the performance is notably enhanced on devices equipped with LiDAR technology, as it allows us to accurately measure the dimensions of detected objects. However, these hardware prerequisites likely will limit our potential user base. Consequently, we are exploring alternative design solutions that are less hardware-intensive with reduced installation footprints

Furthermore, the absence of a comprehensive, readily available database for known products and their features poses a challenge. To address this, we plan to utilize a dual approach: integrating a curated product catalog and supplementing it with crowdsourced data for a more extensive and dynamic collection.

‍

Tools

• Xcode
• IntelliJ
• Java/Spring
• Swift
• SQL (RDS/Postgres)
• NoSQL (DynamoDB))
• AWS (various)
• VisionML
• YOLO
• REST
• Figma

‍

I am a co-founder and the primary developer on Doculist. It has not been launched to market yet, but is still under development and the team is seeking funding to expand development and bring the product to market.