Deep Neural Networks training is handled with the different parameters like size of batch, rate of learning and time need for training. Network training with the parameters decides speed of training the network and it is expected to be low as much as possible. Use of machine learning and artificial intelligence in agriculture increasing day by day. In this research work, the experimentation of training the AlexNet network is done using MATLAB R2020a tool. The investigation of trained network’s training time requires the variable batch size and learning rate consecutively. The behaviour on size of batch on the performance of AlexNet is analysed. The dataset is used in this training of AlexNet is taken from the farming images collected from the Vidarbha region of Maharashtra, a state of India. As we are using more data for analysis so required to use data science techniques for managing more data. The obtained results show the high accuracy always not achieved by keeping high learning rate. The higher learning rate needs more training time to train the network. Not only keeping learning rate low and small batch size will allow training better but it requires more training time additionally. The batch size is affected by another hyper parameters such as learning rate so the combination of these hyper parameters is as important as batch size itself.

Weed Classification, Machine learning, deep learning, AlexNet;