YoloWeb — Streamlit YOLO Object Detection

A simple Streamlit web app that runs YOLO (Ultralytics) object detection on user-uploaded images. Supports multiple models, adjustable confidence/IoU, and works locally or on Streamlit Community Cloud.

Demo App

davidinations-yolo.streamlit.app

Features

• Model selection in the sidebar:
  • Pick a default model from the models/ folder (auto-loaded).
  • Or upload your own .pt model file directly in the UI.
• Sample-image buttons in the sidebar: each button auto-loads a test image from test_images/ that matches the selected default model (e.g. sample1.pt ↔ sample1.png). A button is enabled only when its matching model is selected.
• Image upload: JPG, PNG, BMP.
• Confidence threshold slider and IoU threshold slider.
• Shows the original and annotated image side by side.
• Results table with class, confidence, bounding box, and per-class counts.

Getting started

Install

pip install -r requirements.txt

Add a default model

Place one or more .pt model files into the models/ folder:

models/
  best.pt
  my_model.pt

They will appear in the "default model" dropdown automatically. Without any model in models/, you can still upload a model file from the UI.

Model weights are not committed to git. Add models/*.pt to your .gitignore if using git.

Run locally

streamlit run app.py

Deploy to Streamlit Community Cloud

1. Push the repository to GitHub.
2. Enable Streamlit Community Cloud on the repo.
3. Streamlit reads requirements.txt automatically — no build step.

Configuration defaults

The default confidence and IoU values match the original Flask template's .env:

Parameter	Default
Confidence threshold	0.8
IoU threshold	0.9

Both are adjustable via sliders in the app's sidebar.

Project layout

app.py            # Streamlit application
requirements.txt  # Pinned dependencies
models/           # Optional default .pt model(s)
test_images/      # Optional sample test images matched by model name
README.md         # This file