I need to know whether I should consider the number of labels available or the number of images available to increase the model performance. I mean in my case in a single image I have multiple annotated labels available. I know when I increase the number of images for training the accuracy of the model will go high. But in my case do I need to pay attention to maximizing the number of labels in the dataset or the number of images in the dataset to get better accuracy.
Ex:
Case 1:
1 image contains 4 annotated labels(2 for ClassA , 2 for ClassB)
|
|
Total: 60 images
Case 2:
image_b_1 contains 1 annotated label(1 for ClassA)
image_b_2 contains 1 annotated label(1 for ClassB)
|
|
Total: 200 images
Which case will give the maximum accuracy results during the training?
There are several factors for high accuracy estimates based on the method/model/algorithm at hand. For accuracy and speedy computation, we need appropriate simplified model, sufficient data according to the given model, labeling and good data preprocessing protocol.
There are several factors that affect the accuracy of a model. You need to clearly state the actual context of this problem.
Hritam Basak Muhammad Ali Exactly,I know there are many factors present which can effect to the accuracy .But it's almost clear that for any model if you increase the dateset size , there is a high possibility of getting a better accuracy score .I need to know just ignoring other factors when it comes to the context of dateset size , which of the above mentioned cases will give more accurate results.
For further clarifications:
Imaging I am developing a model to perform instance segmentation of different kind of fish available in a given image.(At most I will try to capture 2 different types of fish, we will consider the types as A and B)
Hence there will be 3 classes present by considering the background.(2+1)
Case1 dataset:
image_1 ----> contains 4 fish(ClassA-2 fish,ClassB-2), Hence number of labels for images will be 4
|
|
Total images = 60
Case2 dataset:
image_1--> contains 2 fish(ClassA- 1,ClassB -1)
|
|
Total images = 75
Now if you compare the datasets you can see Case1 dataset has 4 labels per image * 60 = 240 annotated instances(labels)
Case2 dataset contains 2 labels per image * 75 = 150 annotated instances(labels)
Summary of datasets :
Case1 dataset has less images but more labels
Case2 dataset has more images but less labels
So my question is which dataset will give more accuracy in the context of increasing the dataset size ?
Your approach is OK. I would like to suggest you to focus on the model selection (appropriate statistical model) as well. You can gradually enhance the accuracy with more data availability and labeling. Some clustering model, e.g., von Misses Fisher Distribution on unit hypersphere may be one the initial choice.
Muhammad Ali Which case is more productive?.What's the exact answer.Is it number of labels or images?
- Badges
- Science topic