Hello, I am a beginner to Deep learning! I have just enrolled in the coursera course to learn more about CNN for a project I would like to work on. The project consists of detecting 3 categories of animals say dog, cat, mouse, if the image doesn’t contain any of those, I should somehow say “none”. What would suggest the approach should be? I am confused whether I could do this using Image Classification (Tensorflow + Keras) or I should use Object Detection (Yolo…).
Thank you so much for your feedback.
Yes, if you are doing a “multi-class” classifier like that with a softmax output layer, you would want to have a fourth category for “other” or “none”. Of course this will need to match the labels on your datasets. For example, if you have a general purpose input data set with additional labels that you don’t care about (elephant, kangaroo, alpaca), then you’d want to preprocess those to relabel them as whatever label you choose for “none of the above”.
As to which type of model you use, it depends on what your goal is and how your images are labelled. Can there be more than one type of animal in each image? Do you care only about whether the animal is in the image or do you need to be able to draw a bounding box around it? In the latter case, you are going to need a training dataset that is also labelled that way and you’ll need to use an algorithm like YOLO. If you only care about classifying the images by one type of the labelled animals being present or not, then a simpler classifier is sufficient.
1 Like
Thank you for your help! I only need to know wether the animal is there or not! I guess a simple classifier is enough!