Can you guess what the compressed image might be? It starts out virtually impossible, but as you guess, the image will start to take shape! Use your mouse to scroll through the images in the bottom bar. Then, click on one to make a guess Aim for as few guesses as possible. I can usually get it in 2 or 3 guesses... see if you can do better! Note: Fixed a bug where a black screen would appear with an impossible question... if this occurs for you, please let me know.
Unfortunately, I could only fit 10 images in the project... the original plan was to have 18 (or even more!) I might work on compressing the images down further if people enjoy this. The compression algorithm is low-rank matrix approximation. It is never used for image compression in the real world (there are much better techniques out there, e.g., jpeg DCT), but it is simple enough to quickly implement in scratch. The original project gives a brief description of how it works. Note: While it isn't frequently used for image compression, low rank matrix approximation is related to many other important techniques (e.g., face recognition, principal component analysis, movie recommendations, etc.) through the use of singular value decomposition (SVD) Credits: - All of the compressed images in this project were listed as public domain - Clip art images came from the scratch image library - Used GIMP to scale the images to 128x128 - SVD Decomposition was performed using numpy SVD and matrix values were exported using base python Updates: - Fixed black screen bug (hopefully) - Changed corrupted title to a more sane one To-Do: - Implement low-rank matrix appx for partial SVD (reducing the size of each image greatly) - More images!! (reduce data requirements) - Cloud leaderboard? Or maybe a cloud score distribution? - Mobile support (I don't know what this would entail)
