IceVision Streamlit App

Pre-requisites

You need to have the IceVision package already installed in order to run the IceVision Streamlit App. You can either install the [inference] or [all]packages. The [inference] option is recommended if we are only interested in getting the predictions (inference) as opposed to training models.

For [inference] the packages option:

pip install icevision[inference] icedata

For [all] the packages option:

pip install icevision[all] icedata

Streamlit Local Installation

From your the project root directory , run the following command:

pip install streamlit

In your termilal, press Ctrl+Click on the URL to open the app in your browser, or open the following URL in your browser: http://localhost:8501/.

Drag & Drop an image in the grey zone. The predicted image will appear after a few seconds. Use the controls on the left to adjust the thresholds as well the masks attributes (label, bounding boxes, masks)

---

Running the Streamlit App from the Githup Repo

Install the streamlit package if it has not be done yet by running the following command:

streamlit run https://raw.githubusercontent.com/airctic/icevision-streamlit/master/app.py

In your termilal, press Ctrl+Click on the URL to open the app in your browser, or open the following URL in your browser: http://localhost:8501/

Drag & Drop an image in the grey zone. The predicted image will appear after a few seconds. Use the controls on the left to adjust the thresholds as well the masks attributes (label, bounding boxes, masks)

Running the Streamlit App on a Local Machine

streamlit run app.py

Press Ctrl+Click on the http://192.168.2.11:8501 URL to open the app in your local browser

---

Docker

Building Docker Image

You might choose your own tag name instead of app:latest

docker build -f Dockerfile -t ice-st:latest .

Running the Docker Image

docker run -p 8501:8501 ice-st:latest

Important: Your container will be available on http://localhost:8501/

Note: Make sure no other app is running on port 8501

Drag & Drop an image in the grey zone. The predicted image will appear after a few seconds. Use the controls on the left to adjust the thresholds as well the masks attributes (label, bounding boxes, masks)