You are browsing as a guest. Sign up (or log in) to start making projects!

CoinCounter

By @fast

  • 2 Devlogs
  • 65 Total hours

App to make counting money easier.
Ship #1 Changes requested

@fast on CoinCounter

I built Coin-Counter, an app that lets you upload a picture of your loose change or bills and counts them instantly. It supports USD, EUR, and PLN currencies. The hardest part of the entire project was easily the dataset, finding, cleaning, and combining images for all these different coins and banknotes was a massive headache. But after training the model on over 100k images for 30+ hours across multiple attempts, it actually works! The web app has a clean, responsive UI with interactive sliders to tweak detection sensitivity. You can run the entire project easily using Docker Compose, just make sure you have Git LFS installed on your system before cloning so the actual model file downloads correctly!

  • 2 devlogs
  • 65h
Try project → See source code →
Open comments for this post

@fast on CoinCounter

51h 57m 23s logged

Okay, it’s been a while…

I’ve been struggling with this project for a long time. First, cleaning up a massive amount of data took ages. By the end, I had around 100k images in total! Then came the training… and honestly, ugh. Constant tweaks and trial-and-error almost drove me insane. The total training time is surely 100+ hours, considering all the restarts and reruns.

But finally, I’ve got something like a working model, and that’s what I’m sharing now. I’ve also finished the frontend, and everything is finally available for you to try. It’s also open-source (though I couldn’t share the dataset due to uncertain sources and potential legal concerns).

Any feedback is greatly appreciated!

@fast

Okay, it’s been a while…

I’ve been struggling with this project for a long time. First, cleaning up a massive amount of data took ages. By the end, I had around 100k images in total! Then came the training… and honestly, ugh. Constant tweaks and trial-and-error almost drove me insane. The total training time is surely 100+ hours, considering all the restarts and reruns.

But finally, I’ve got something like a working model, and that’s what I’m sharing now. I’ve also finished the frontend, and everything is finally available for you to try. It’s also open-source (though I couldn’t share the dataset due to uncertain sources and potential legal concerns).

Any feedback is greatly appreciated!

Replying to @fast
0
22
Open comments for this post

@fast on CoinCounter

12h 34m 6s logged

So I’m building an app that counts money from a photo, but honestly, local training has been a total nightmare. Spent 12 hours straight just cleaning and stitching datasets together, only for training on my 3070 to go horribly because of VRAM limits forcing me to downscale everything. Looking at the test image, you can see it’s detecting something, but it completely misses the fine details and fails to classify them because it has zero concept of relative sizes. I’m pivoting to a Transformer-based model (RF-DETR) so it can actually look at the whole picture and compare coin sizes better. Renting a beefy GPU in the cloud to do the training.🙄

@fast

So I’m building an app that counts money from a photo, but honestly, local training has been a total nightmare. Spent 12 hours straight just cleaning and stitching datasets together, only for training on my 3070 to go horribly because of VRAM limits forcing me to downscale everything. Looking at the test image, you can see it’s detecting something, but it completely misses the fine details and fails to classify them because it has zero concept of relative sizes. I’m pivoting to a Transformer-based model (RF-DETR) so it can actually look at the whole picture and compare coin sizes better. Renting a beefy GPU in the cloud to do the training.🙄

Replying to @fast
0
7

Followers

Loading…