Am completely a Novice in Artificial intelligence and my final year project based on training a model.
My topic is Early Detection of River Blindness using machine learning.
Problem: the onchocerciasis which also refers to as the river blindness is a disease gotten as a result of blackfliies bites on the body which causes some effects that could turn to blindness.
Proposed solutions:
I want to build a machine learning program that will detect not predict if someone has been infected or not.
I want to train a model with at least 2000 images and the model should be able to accept image from the users in order to detect if the person has river blindness or not.
If you have formal ML knowledge then this is a simple task, they way you mean detect are you pointing to object detection?
After the knowledge you have I would recommend first check for bias in data (imbalanced dataset)
Use a ANN as Baseline see how they work, to predict ( yes predict, it is the first step before you detect)
Then you can play around with YOLO, Resnet, Inception
If you are comfortable till now Try ViT Transformer for object detection
It is very interesting topic. Since you are using image data, I would suggest you to review CNN (convolution neural network). There are many short courses available that can get you basic knowledge. But you need some language skills to implement this e.g python. You can also check MATLAB.
For image data, you can search on Github whereby folks could have train the model with images related to what you are having. You reuse the model and use Transfer Learning technique to freeze the initial layers and train the last few layers with your own images etc. How many layers to freeze will depend on the problem. You can watch Prof Andrew Ng videos on CNN and Transfer Learning for more details. Give it a shot and pose your questions if you need further assistance.