velocity - Stardance

Ship ✨ Blessed

This is my second ship.

For my first ship, I had built a very barebones app with a lot of bugs and issues. It simply turned images into triangles, and there was a lot of room for improvement.

For this ship, I tried to implement everything that the reviews had asked for. I added a gallery, expanded the README significantly, made my devlogs more detailed, and added clear instructions explaining what each parameter does.

I’m very happy with how this project turned out, especially considering how much I improved it from my first ship.

The only major improvement I would still like to make is getting the program to run locally. That would be a challenge because of the limited resources available in a browser, but it’s something I’d like to work toward in the future.

Since I have a lot of pictures I am going to upload one of my favorite ones, a photo I processed about 3 hours into this project.

3 devlogs
2h
12.84x multiplier
36 Stardust

Try project → See source code →

Open comments for this post

@velocity on TRIANGULATOR · 10 days ago

1h 0m 15s logged

Final post for this project :<

I have finally added a processing screen so you know when the images are actually processing. I have added a gallery of my favorite outputs yet. Added instruction on how to use the website and cleaned up the UI, Check out the website

Open comments for this post

@velocity on TRIANGULATOR · 10 days ago

1h 0m 1s logged

I cleaned all the code, fixed a massive deployment error (sadly the time did not count). I significantly improved my README.md, Read my README.

Open comments for this post

@velocity on Frision · 13 days ago

2h 5m 36s logged

Added an Object classifier
Once the object detector detects the images I use pillow to crop into them and use a classifier with 900 items to classify that food it has really good accuracy, an average photo of a fridge has about 50+ items

Open comments for this post

@velocity on Frision · 13 days ago

1h 1m 52s logged

Breakthrough in my project :)
I am finally able to detect a lot of objects clustered together easily, now the next phase is actually checking what each of those objects are

Open comments for this post

@velocity on Frision · 18 days ago

1h 45m 35s logged

Managed to figure out how to get a Computer Vision Model Working on my computer

after ages of searching for pretrained models that are fast, free and can detect a large number of classes. I found a YouTube video explaining Grounding Dino. I think it might be the savior of all of computer vision or at least my project, i managed to install the hugging face version of it to be able to analyze images of fridges and list out all items currently in it. Had to find a list of foods which I found on USDA as there common grocery foods. Around 900 of the most common foods, I tried using AI models, such as qwen or gemini, but they took way to long and burned through tokens.

Right now Grounding Dino is not able to handle 900 inputs so I will have to figure out how to batch the classes.

Open comments for this post

@velocity on TRIANGULATOR · 27 days ago

21m 37s logged

Fixed my triangulator for transparent images

Ship

@velocity on TRIANGULATOR · 27 days ago

I have made a simple web app that takes in and image a turns it into triangles, hence the name Triangulator. I am very proud of the algorithm i have written myself for colour detection, placing points in backgrounds and finding midpoint of each triangle. This also helped me revisit some html and css. I would love for people to play around with my project and in the future I might make my project better. Something I still want to improve is time as vercel servers are slow, so the large number of calculations takes a lot of time.