Hello fellow researchers! I am currently doing my final year project which involves Image Recognition with a Supervised Learning Machine Learning Algorithm. Currently, I have datasets comprised of images (without labels of course) that are obtained from Kaggle. The project that is in progress is to inspect the quality of the green coffee bean and classify them to be either healthy or defected. I am new to this field per se, and I find it hard to do the Python and to train my model. Am I in the right course to train my model with just images? Isn't it like unsupervised learning, since my current datasets are labelled but without their input variable.
I think the question is not really clear to me. First, you say that you are working on a project that aims to classify the quality of coffee beams into two categories using images, right? Then, that is a binary classification task and it is supervised learning because you have the labels (i.e. healthy or defected) and you have the images (the inputs). Something like a CNN would process the image and output a single value between [0, 1], which is compared to the target (label).
As a reference, I think this paper (https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8090980) is doing what you want.
Giorgio Morales thank you for replying, and I'm sorry that is not clear to you, as mentioned before I am new to this field and quite confused with the terminology here and there. So does that mean that I can train my models by just using images? So far from what I have researched, many Machine Learning models are using csv. file format to train their model (correct me if I'm wrong). And for this project, I am using SVM, kNN, and Naive Bayes classifier (because one of the project's objective is to evaluate the accuracy among the models).
Shafian Izan, you can use any type of input format that you want. In this case, the natural input to your model is an image so there are many ways in which the data can be stored, such as .jpg or .tif files, or the data could be already embedded in a .h5 file. When you say that many datasets are in CSV format is because they use tabular information, which is not your case.
Since you're using SVMs, kNNs, and Naive Bayes, those models require one-dimensional inputs, which means you need to process your images and reshape them into just one dimension, that would be your input. In other words, say that you have 100 images of 32x32 pixels, you have to read them all from their files (whatever the format is) and store them (using numpy maybe) in a matrix of size (100, 32, 32, 3) (I am assuming your images are RGB); then, you need to reshape the matrix to a new dimension (100, 32x32x3).
There are several sources to obtain data from. Note that dataset can be in different format to meet ML requirements.
However, you can always convert data from one format to the other using various tools. Having this at the back of your bind will expand your searching data availability scope.
Check these links for possible sources:
https://www.kaggle.com/datasets
https://datahub.io/machine-learning
https://lionbridge.ai/datasets/the-50-best-free-datasets-for-machine-learning/
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training