Openmv implements target detection through IMPULSE training model

Openmv neural network

foreword

openmv4plus can carry out neural network learning and realize target detection. It has learned K210 target detection and machine learning for target detection before. Use openmv4plus for object detection today.

I have done a lot of understanding before learning, openmv is really convenient for target detection.

1. Cloud training

URL: IMPULSE URL

2. Operation steps

1. Collection of data sets

Create a new folder to store the data set.

Create a new folder and enter the category name


to collect the data set. After the collection is completed, the corresponding folder will display the data set

2. Upload dataset

It can be uploaded through the API key, and the computer firewall needs to be turned off, otherwise an error will be prompted. Here is another upload method.

First select Data acquisition, click the arrow to the left of Collected data (Upload existing data).

Select the corresponding file (Ctrl+A to select all)

, select Enter label, and enter the name of the category to be classified (note that the training model must have two or more categories).

If there are many categories, upload them one at a time.

3. Training model

1. Create a model

Click Impulse design - Create impulse, and select the configuration according to the figure

2. Parameter generation

3. Start training

These settings can be set by default. Of course, the more learning rounds, the better. During the training process, the IMPULSE platform will automatically stop the training to prevent overfitting if the training effect is good.

4. Model testing

5. Export the model

6. Effect test

Copy the exported file to openmv and

run the PY file

Summarize

The neural network training model using Openmv is probably these, and the method is very simple.