☠️ DetoxifAI

Detecting venomous or poisonous species in the wild

👀 What is it?

DetoxifAI's main goal is protect people who arent knowledgable about the species they may encounter. For exmaple if you were walking in the woods and you wanted to check if a mushroom was edible or not, you could easily use DetoxifAI to find information and said species. Currently it focuses on commonly found California snake and mushrooms species.

⚙️ How does it work?

DetoxifAI uses the YoloV5 computer vision model to classify certain species using images uploaded by the user. Our AI was created like so:

Web scraping images of specific species of mushroom and snake we wanted to classify using serp
Label and filter through data found so we can get the images we will use for training our model
Export our data and augment it so that it creates a stronger final model
Train the model using Google Colab and Kaggle for superior speed GPUs. Adjusting hyperparameters and input data accordingly
Create and frontend and also a backed using flask for serving it to users, while also linking up our backend to YoloV5 and classifying each image uploaded by a user

👏 The Team

Jinxuan: Professor
Chase: Instrucor
Toryn: Frontend Developer (me!)
Michael: Backend Developer
Arhant: Machine Learning Engineer
Ayush: Data Scientist
Sravya: Data Scientist
Jacky: Product Manager

🚀 Usage

To use DetoxifAI first clone this repo by running

git clone https://github.com/zahtec/SC22-BatchB-Flaming-Fuelers-DetoxifAI.git

Then change your terminal directory to the SC22-BatchB-Flaming-Fuelers-DetoxifAI folder and run (assuming you already have Python installed)

python ./app/main.py

Or, if you have multiple versions of python installed

python3 ./app/main.py

This will intiate a flask server that will bind the URL to localhost:1234. Open your browser, go to this URL, and proceed to upload your images & get your results!

This project is also able to run on CoCalc and has been tested on there

📈 Stats

More information can be found on this projects about page on the frontend, if you would like to view it, please follow the instructions above (🚀 Usage)

Final precision and recall

Precision: ~0.97

Recall: ~0.98

Confusion Matrix and Roboflow image totals

🙂 Thanks for viewing!

This was made at AI Camp. Where seers become beleivers, get a chance to participate in a super cool camp and get a possible internship at AI Camp, linked above!