Solar Panel Rotator

This app, built with streamlit, performs simple rotate operation on input images. 😎 The demo illustrates the whole steps of solar panel orientation prediction and also contains the function of labeling.

Steps of Orientation Prediction

This section mainly show the main steps of panels orientation prediction. There are four main steps, including:

Step 1: Predict solar panel.

At first, solar panel can be detected through existing models such as YOLO, Faster RCNN. Our solar panel detector only predicts the orientation of panel based on the detected results instead of detecting solar panels.

Step 2: Cut the target from the image.

Crop the pictures to facilitate the prediction of the single small solar panel in the back.

Step 3: Predict rotation angle.

Input the cropped images to the model, and then output prediction angle.

Step 4: Adjust rotation angle and get the orientation of panels.

In order to get the orientation of panel, we need to add or subtract 90 degrees from the prediction angle, the output of step 3.

Dependencies

This code uses the following libraries

• python 3.7+
• streamlit
• numpy
• matplotlib
• pillow
• opencv-python

Run

Run the code below to run the local version of Solar Panel Rotator.

streamlit run main.py