I’m trying to tokenize some images to train a HuggingFace model. In the process, I dowloaded and stored locally the images. The path to each of these images have been stored in a dictionnary of lists, in which the key “image_path” contains a list of path (strings) such as “.mypathto.jpeg”. The next step is to use these path one at a time to process these images.

When I’m trying to loop into the dictionnary and the lists, nothing seems to happen. The only error I get is from my try-exception which gives me error found : image_data. Here’s the dict :

self.dataset = {
            "image_path": [],
            "label_path": [],
            "image_data": [],
            "label_data": []
        }

Here’s my code :

def images_to_tensor(self):
        """
        Convert images data to pytorch tensor by converting them to RGB values and then tensors using model processor.
        
        Parameters :
        None

        Returns:
        None
        """
        
        try:
            for key, value_list in self.dataset.items():
                for value in value_list:
                    if key == "image_path":
                        with open(value, "rb") as file:                    
                            image_rgb = Image.open(file).convert("RGB")
                            image_tensor = self.processor(images=image_rgb, return_tensors="pt").pixel_values.squeeze(0)
                            self.dataset["image_data"].append(image_tensor)
       
            print(f"dataset length image_data : {len(self.dataset["image_data"])}")
        
        except Exception as e:
            print(f"error found : {e}")
        
        return None

I also tried with a list of dictionnaries but got the same problem.