src/FaRL.py

#import torch
#import clip


#class CLIPImageEncoder:
#    def __init__(self, device="cpu"):
#        self.device = device
#        self.model, self.preprocess = clip.load("ViT-B/16", device=device)
#
#    def encode_image(self, image_pil):
#        print("Encoding image with CLIP")
#        with torch.no_grad():
#            image_preprocessed = self.preprocess(image_pil).unsqueeze(0).to(self.device)
#            image_features = self.model.encode_image(image_preprocessed)
#            return image_features.cpu().numpy()[0]