I read some blogposts that Resnet50 can be used to extract features from images. But I am not sure if the vector representation obtained from this model will be a good descriptor of an image. Is there any other model which can give better embeddings for an image?
Zabir Al Nazi
Yes, you can use pre-trained models to extract features. You can study the feature performance from multiple models like vgg16, vgg19, xception, resnet-50 etc. and do a comparison. You can look into my notebook where I have extracted features for melanoma classification. https://github.com/zabir-nabil/lesion-segmentation-melanoma-tl/blob/master/PH2_TL_classification.ipynb
0 votes
0 thanks
Omid Mohammadi
Yes,Certainly this is easy to do. See Chollet's Deep Learning Book for the full tutorial.
0 votes
0 thanks
- Badges
- Science topic
- Similar topics
- Computer Science
- Images
More Ekta Raj's questions See All
Similar questions and discussions