Image Segmentation with Unsupervised and Pretrained Models

Here in this repo we try to segment some images using 3 types of methods:

Unsupervised Segmentation

Segment all images using the following algorithms:

• Fuzzy C-Means
• K-Means

Feature-Based Segmentation

Using a pretrained ResNet model (or any other suitable method):

• Extract feature maps from the input images.
• Resize these maps using interpolation or another technique to match the original image size.
• Concatenate the resized feature maps with the original image as input.
• Apply Fuzzy C-Means and K-Means segmentation again using this augmented input.

Segmentation with SAM Model

Use the pretrained SAM (Segment Anything Model) to perform image segmentation on the same dataset.

Results