ESPWandPy

This is the recognition algorithm part of the wand

The random forest algorithm is used

I have tried using neural networks/deep learning but found that the recognition performance is extremely poor (possibly due to insufficient data or code issues)

Data collection data_comllection-server. py

Send data to: 5000/upload_data (you can modify the ESP code and send it directly here)

Model training train_madelB.py

Just run it directly

Identify predictioneserverB.py

POST data to: 5000/advance

The result returns the key value pairs of the probability of the corresponding action (can be modified to directly return the maximum value in the comments)

ESPWandPy

这是魔杖的识别算法部分

使用的是随机森林算法

本人尝试过使用神经网络/深度学习 但是发现识别效果极差(可能是我数据量不够或者代码有问题)

数据收集 data_collection_server.py

数据发送到:5000/upload_data (可以修改ESP的代码直接发到这)

模型训练 train_modelB.py

直接运行即可

识别 prediction_serverB.py

POST数据到:5000/predict

结果返回对应动作的概率的键值对(可以修改直接返回最大值 在注释)