Stranger Sections - Winners Announcement!

Thank you to all those who participated in the Stranger Sections Challenge! This was a new style of challenge for Xeek and as always our amazing community showed up and hit it out of the park! We received several high-quality submissions, each with a diverse methodology for solving the problem. To provide more transparency into the evaluations, we wanted to provide a summary of the challenge winners. Please feel free to reply to this post and share/discuss your approach with the community!

Now for the winners:

Best Overall Model is awarded to Igor Ivanov. Igor used Meta’s Segment Anything Model (SAM) with a Zero Shot Transfer approach to segment the images. Igor not only provided the best segments, but also provided the functionality for users to experiment with segment size and predicted IOU cutoffs. Igor’s submission was well organized and included great markdown discussion which helped the judges understand the process. Great work!

Best Segmentation Model is awarded to Ramdhan Wibawa. Ramdhan’s submission uses a CUTS model. This is a state of the art tool for unsupervised segmentation of medical images. This was a unique approach that was well documented. Great work!

Best Transfer Learning & Clustering Models are awarded to Daniel B. Daniel’s submission was developed from a differentiable feature clustering paper that uses ResNet50 to extract features from the images, then KMeans to cluster those features. Finally, Daniel builds a neural network to segment the images in each cluster. This was an innovative approach that was well documented and allowed judges to experiment with the clustering and segmentation hyper-parameters. Great work!

Best Visualizations/Deployment is awarded to Evgeny K. Evgeny’s submission provided great image analysis and visualizations. Evgeny not only displayed image segments, but they also provided an exploratory data analysis of images, including an analysis of color schemes via histograms and an analysis of images in three channels of gray and HSV. Great work!

Thank you again to all of the challengers. Be on the look out for Stranger Sections Part 2 coming next year!