Datasetpublic

Man1103/PetImages

This is clone of popular Cats Vs. Dogs image classification dataset for CNN-based visual classification models. If you are using raw images (stored in /data), this script can be used for creating train-test split on the raw images dataset: def fetching_selected_data(cats_images_subpath: str, dogs_images_subpath: str, test_split: int, max_len: int = 1000000000): total_len = min(len(os.listdir(cats_images_subpath))… See the full description on the dataset page: https://huggingface.co/datasets/Man1103/PetImages.

sourceHugging Faceupdated 22h agoView on Hugging Face

1likes43downloads

Dataset Card

This is clone of popular Cats Vs. Dogs image classification dataset for CNN-based visual classification models.

If you are using raw images (stored in `/data`), this script can be used for creating train-test split on the raw images dataset:

def fetching_selected_data(cats_images_subpath: str, dogs_images_subpath: str, test_split: int, 
                           max_len: int = 1000000000):
    total_len = min(len(os.listdir(cats_images_subpath)), len(os.listdir(dogs_images_subpath)), max_len)

    training_set_count = int(total_len * (1 - (test_split/100)))
    testing_set_count = total_len - training_set_count

    def _get_image_files(parent_dir: str, first_count: int, max_count: int):
        all_files = []
        count = 0
        for child in os.listdir(parent_dir):
            grandchild = os.path.join(parent_dir, child)
            if not os.path.isdir(grandchild):
                count += 1
                if os.path.isdir(grandchild):
                    continue
                final_full_path = os.path.join(parent_dir, grandchild)
                all_files.append(final_full_path)
            else:
                _get_image_files(grandchild)

        last_count = first_count + max_count
        return all_files[first_count:last_count]


    def _get_input_labels(first_count: int, max_count: int):
        cats_list = _get_image_files(cats_images_subpath, first_count=first_count, max_count=max_count)
        dogs_list = _get_image_files(dogs_images_subpath, first_count=first_count, max_count=max_count)
        img_files_list = cats_list + dogs_list

        X, y = [], []
        for file_path in img_files_list:
            img_file_data = Image.open(file_path).convert("RGB")
            X.append(img_file_data)

            if 'cat' in file_path.lower():
                y.append(0)
            elif 'dog' in file_path.lower():
                y.append(1)

        return X, y

    train_first_count = 0
    train_max_count = training_set_count
    X_train, y_train = _get_input_labels(train_first_count, train_max_count)

    test_first_count = training_set_count
    test_max_count = testing_set_count
    X_test, y_test = _get_input_labels(test_first_count, test_max_count)

    return X_train, X_test, y_train, y_test

If you are using tensor format data:

—Please note that, these tensor datasets are not fully comprising of >24k images of cats and dogs.
—Due to compute limits, a lesser subset of dataset (for example., 10k from cats and 10k from dogs) is used to create these safetensor datasets.
—Training data: Xtrain, ytrain pairs in /train.safetensor
—Testing data: Xtest, ytest pairs in /test.safetensor