It should work in any compatable python version.
You could run TensorFlow for CPU Only ( or/and GPU).
kaggle repository (+augmented) image dataset and .npy files https://www.kaggle.com/georgiosgiouvanis/poacher-detection-3classes
Version Used
python : 3.8.5
numpy : 1.18.5
Keras : 2.4.0
tensorflow : 2.3.0
Use your own images: You will find code to resize images to prefered dimentions, also code to augment your data.
step 1: Create virstual inviroment virtual environment. Install miniconda+keras tensorflow https://www.youtube.com/watch?v=PnK1jO2kXOQ&t=884s
step 2:Install keras tensorflow https://github.com/jeffheaton/t81_558_deep_learning/tree/master/install
step 3:Install all additional libraries
step 4:Prepare dataset , put files(=number of classes) with images to your working directory the code will label the three different classes and prepare the input for the network.
step 5:Model(Convolutional Neural Network) , Choose ,Train, Evaluate.
Poacher detection with 3 classes.
Classes 3: 2-Poachers with guns, 1-Poachers with arrows, 0-No Poachers.
The keras model accieved 90% accuracy in 20 epochs
The images collected manualy resized to 224x224 (3collors) and augmented.