PicMyPic

Inspiration --

After we, as a group, took trips to places like Florida and Tennessee, we’d come back with way too many photos and spend hours going through them trying to "pic" and choose which ones were actually worth keeping. Most of them were nearly identical with small differences. PicMyPic was built to solve this problem. Instead of manually going through your camera roll, we wanted a system that could automatically rank your photos, saving you precious time to create more memories.

What it does --

PicMyPic lets users upload a batch of photos and automatically ranks them from best to worst. The system first determines whether the image is a single subject, group photo, or landscape, then scores it based on important features like eye openness, sharpness, expression, exposure, composition, and overall image quality. Each photo receives a weighted score and is sorted so users can instantly see which pictures stand out.

How we built it --

We built PicMyPic in Python using computer vision libraries to extract measurable image features. Our pipeline handles image ingestion, face detection, head count classification, feature extraction, and weighted scoring. The metrics we use contain things like Laplacian variance for sharpness and brightness analysis for exposure, and we also made sure the system supports both JPEG and RAW files. The core of the project was designing and tuning the scoring logic to reflect real photography principles.

Challenges we ran into --

Tuning the scoring weights was by far our biggest challenge. Early on, brightness was being prioritized too heavily. This causes overly bright images to rank higher than sharper, better photos. We had to iterate several times to balance our logic. We also dealt with invalid image errors, cross-platform inconsistencies, and time pressure while implementing RAW file support, while trying to improve overall accuracy.

Accomplishments that we're proud of --

In 48 hours, we built a full end-to-end system that classifies, scores, and ranks photos automatically. We’re very proud of the structured scoring algorithm and how it adapts to different types of images. Looking back at it, seeing the system consistently push the best photos to the top made all the debugging and tuning worth it.

What we learned --

This project strengthened our experience in computer vision, feature engineering, and algorithm tuning under tight deadlines. We also learned how important clear role assignment and team coordination are in a hackathon environment. Balancing speed with accuracy was one of the biggest takeaways.

What's next for PicMyPic --

Next, we want to improve accuracy with a larger and more diverse dataset, allow bigger batch uploads, and potentially build a mobile app that can access a user’s photo library directly. Long term, we see PicMyPic becoming a smart filter that instantly points out your best photos as soon as you take them.