DinoV3 for image matching
#3
by
sagaragrawal
- opened
I want to use the DINOv3 model for a fine-grained image matching task. Given a pair of e-commerce product images, the model should predict whether the items match. The challenge is that product images may have very subtle differences, varying orientations, or even partial visibility in one image compared to full visibility in another. Could someone guide me on how to approach this problem? Will this new model be able to capture such fine grained differences?
for example below images -
