https://github.com/moonshallow5/food_vision_mini
A program which can detect if an image contains either a pizza, sushi or steak: Has now been deployed on Flutter :)
https://github.com/moonshallow5/food_vision_mini
huggingface matplotlib numpy pytorch
Last synced: 8 months ago
JSON representation
A program which can detect if an image contains either a pizza, sushi or steak: Has now been deployed on Flutter :)
- Host: GitHub
- URL: https://github.com/moonshallow5/food_vision_mini
- Owner: Moonshallow5
- Created: 2024-09-03T19:10:53.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-09-22T03:46:56.000Z (about 1 year ago)
- Last Synced: 2025-01-12T09:30:42.964Z (10 months ago)
- Topics: huggingface, matplotlib, numpy, pytorch
- Language: Python
- Homepage:
- Size: 58.9 MB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Food_Vision_mini
## 💠What this project does
I wanted to have a more hands-on experience in machine learning, so I decided to make a food image classification program. My program can take an image of either a "sushi", "pizza" or a "steak" and would be able to return a conclusion on what that image actually is
## 💡 What I implemented
In this program I implemented multiple different models to test which model has the highest accuracy in classifying if an image contains either a sushi, pizza or steak.
My test results are shown on results.txt, where I used pre-trained pytorch models of efficinentnetb2 and resnet50, as well as my own model which has a much lower accuracy.
Furthermore, I also implemented my own custom implementation in model_architecture.py, where a TinyVGG and VGG16 is implemented, which has much lower test accuracy as it isn't a pre-trained model and i have a small selection of datasets
Here are Effnet_b2 and resnet50 test results on 7 epochs without any data augmentation
Learn about MLflow as well as Nsight systems to optimize my program and see for any potential bottlenecks as well
## 👀Live preview
I wanted to show my project in live-action so I made a link to my HuggingFace account where there is already an app running on gradio for anyone to use. The model used in HuggingFace and in my Flutter project are models which were trained in this repo.
## 🔧Further implementations
- [x] I want to turn this program into a Flutter API for anyone to use it on their mobile phones, but this would take me some time to lern Dart documentations
