In recent times, the state of agriculture and the amount of work people need to put in to check if plants/food is growing correctly is phenomenal, because it is 2019 and workers still need to organize and recognize the difference between different plants and weeds. People who are working in the agriculture field still have to have the ability to sort and recognize different plants and weeds, which does take a lot of time and a lot of effort in the long term. This trillion dollar industry can be greatly impacted from weeds if action is not taken to reduce and completely despair of it.
This is where Artificial Intelligence can actually help benefit those workers, as the time and energy to identify plant seedlings will be much shortened. The ability to do so effectively can mean better crop yields and in the long term will result in better care for the environment. As by identifying the difference among the plants and weeds in a timely manner where it is highly accurate can positively impact agriculture.
Trending AI Articles:
1. How I used machine learning as inspiration for physical paintings
2. Keras Cheat Sheet: Neural Networks in Python
3. Making a Simple Neural Network
4. Basics of Neural Network
Although the issue of identifying weeds from plant seedlings may not seem concerning, it actually can be, as if weeds are left there with the other plants or misidentified to instead be a plant, in the long term, weeds can bring plants to not grow as much as they do consume a portion of their nutrients. This would largely impact agriculture in a negative aspect as plants will not grow to their fullest and there will be less of them, although it depends on the type of weed that is present there is no point in taking a chance. As depending on the type, weeds can cause crop yield loss, taint food and feed crops, can harbor problem insects and crop diseases, and much can also cause many other problems that would impact the harvest. Weeds typically sometimes look very green and identical to plants, which is why it is sometimes difficult to identify 100% accurately from a human perspective. It is estimated that weeds cost $2.5 billion a year in lost agricultural production within Australia which is a tremendous loss of money and agriculture. Weeds are indeed a growing issue within the agriculture industry.
In order to identify the different plant seedlings, and thus find what are the weeds, machine learning can greatly help the problem of classifying plants and weeds. As by creating a CNN model that can identify the images to place them in the correct category with a high accuracy rate, can ultimately change how weeds are negatively impacting the current state of agriculture.
The following steps were taken in order to create the program on Google Colab that identified basic plant seedling species from weeds.
- In this particular model, the dataset was provided by kaggle which can be downloaded here. The first step that was taken was to analyze the dataset, in order to understand the complexity of the dataset itself. The data set contains 960 unique plants belonging to 12 species at several growth stages. The training and testing data set usually should be 70%-90% train and 30%-10% test. The following are the 12 classes/categories in which the dataset images had to fit in:
2. Load the data and import libraries
3. Organize the images and then use Mask to remove the background of images, this technique to remove the background images appeared to be helpful for isolating the image of the seedlings
4. Prepare the Data, the process for preparing the data included, normalization, encoding, and categorizing the labels from the folders.
5. Training and Validation sets, use 90% for training set and 10% for test sets
6. Data Generator is utilized in order to prevent overfitting. Data augmentation is configured in the image generator which will randomly rotate, shift, zoom and flip image during the fitting of the model.
7. Initialize the CNN model which contain the following layers:
8. Visualize the results in order to check which class has the best and worst performances and necessary steps can be taken in order to improve the results. Seek the accuracy rate in order to find that the validation accuracies and losses converge nicely. The results of model_A within the program are a validation accuracy of 87% and an error rate of 13%.
- Artificial Intelligence can impact numerous industries and ways that are unimaginable and can impact the world such a positive way
- Using CNN’s and actually implementing it within agriculture can lower costs and help save plants thus produce more food