### Install Repository and Dependencies

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Install the repository and its dependencies using conda and pip.

```bash
conda env create -f environment.yaml
conda activate taming
pip install -e .
```

--------------------------------

### Install CLIP and Dependencies

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/README.md

Installs PyTorch, torchvision, and CLIP with necessary dependencies. Ensure to select the correct CUDA toolkit version or use 'cpuonly' if no GPU is available.

```bash
$ conda install --yes -c pytorch pytorch=1.7.1 torchvision cudatoolkit=11.0
$ pip install ftfy regex tqdm
$ pip install git+https://github.com/openai/CLIP.git
```

--------------------------------

### Run Streamlit Demo for ADE20k

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

This command launches the ADE20k demo using example segmentation maps. The --ignore_base_data argument configures the data module for example usage.

```bash
streamlit run scripts/sample_conditional.py -- -r logs/2020-11-20T21-45-44_ade20k_transformer/ --ignore_base_data data="{target: main.DataModuleFromConfig, params: {batch_size: 1, validation: {target: taming.data.ade20k.Examples}}}"
```

--------------------------------

### Create and Activate Conda Environment

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Use these commands to create a new conda environment from the 'environment.yaml' file and then activate it. This is the initial setup step for the project.

```bash
conda env create -f environment.yaml
conda activate taming
```

--------------------------------

### Run Streamlit Demo for D-RIN with Example Data

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

This command runs the D-RIN demo using example depth maps included in the repository. The --ignore_base_data argument specifies the data module configuration.

```bash
streamlit run scripts/sample_conditional.py -- -r logs/2020-11-20T12-54-32_drin_transformer/ --ignore_base_data data="{target: main.DataModuleFromConfig, params: {batch_size: 1, validation: {target: taming.data.imagenet.DRINExamples}}}"
```

--------------------------------

### Run Streamlit Demo for COCO

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Execute this command to run the COCO demo using example segmentation maps from the repository. Ensure the model folder is correctly specified.

```bash
streamlit run scripts/sample_conditional.py -- -r logs/2021-01-20T16-04-20_coco_transformer/ --ignore_base_data data="{target: main.DataModuleFromConfig, params: {batch_size: 1, validation: {target: taming.data.coco.Examples}}}"
```

--------------------------------

### Image to Image Inference Command

Source: https://github.com/biomap-research/morphdiff/blob/main/README.md

Example command for image-to-image generation using gene expression data. Verify input paths and model configurations.

```shell
python MorphDiff/scripts/gene_img2img_5c.py \
  --init_img_path data/drug_base_demo.npy \
  --gene_path data/gene_expression_demo.npy \
  --outdir ../results/drug_inference_demo/ \
  --output_name ../results/drug_inference_demo.npy \
  --model_path MorphDiff/checkpoints/drug_ldm.ckpt \
  --config_path MorphDiff/configs/ldm/morph_5c.yaml \
  --scale 1.0 \
  --strength 0.55 \
  --seed 10
```

--------------------------------

### CLEVRCounts Class and Template Prompts

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Defines the class names and text templates for the CLEVRCounts dataset. Use these to guide CLIP's understanding of object counts.

```bash
classes = [
    '10',
    '3',
    '4',
    '5',
    '6',
    '7',
    '8',
    '9',
]

templates = [
    'a photo of {} objects.',
]
```

--------------------------------

### Sample ImageNet Model for Specific Classes

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

To sample from specific ImageNet classes, provide their IDs using the --classes argument, separated by commas. This example samples ostriches, border collies, and whiskey jugs.

```bash
python scripts/sample_fast.py -r logs/2021-04-03T19-39-50_cin_transformer/ -n 50 -k 600 -t 1.0 -p 0.92 --batch_size 25 --classes 9,232,901
```

--------------------------------

### ImageNet Data Structure Example

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

This illustrates the expected directory structure for ImageNet data when placed in the cache directory. Ensure subfolders for train/validation splits and JPEG images are present.

```text
$
XD G_CACHE/autoencoders/data/ILSVRC2012_{split}/data/
├── n01440764
│   ├── n01440764_10026.JPEG
│   ├── n01440764_10027.JPEG
│   ├── ...
├── n01443537
│   ├── n01443537_10007.JPEG
│   ├── n01443537_10014.JPEG
│   ├── ...
├── ...

```

--------------------------------

### Gene to Image Inference Command

Source: https://github.com/biomap-research/morphdiff/blob/main/README.md

Example command for generating images from gene expression data. Ensure all paths and parameters are correctly set.

```shell
python gene2img_dpm.py \
  --gene_path data/gene_expression.npy \
  --outdir result/g2i/ \
  --output_name /result/g2i.npy \
  --H 128 \
  --W 128 \
  --model_path /model/drug_ldm.ckpt \
  --config_path config/morph_5c.yaml \
  --scale 1.0 \
  --batch_size 2 \
  --ddim_steps 100
```

--------------------------------

### Download and Extract Country211 Dataset

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/country211.md

Use this command to download the 11GB archive of the Country211 dataset and extract its contents into a 'country211' subdirectory. Ensure you have sufficient disk space and `wget` and `tar` utilities installed.

```bash
wget https://openaipublic.azureedge.net/clip/data/country211.tgz
tar zxvf country211.tgz
```

--------------------------------

### Run Streamlit Demo for FacesHQ

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Launch the Streamlit demo for the FacesHQ model by providing the path to the downloaded model folder.

```bash
streamlit run scripts/sample_conditional.py -- -r logs/2020-11-13T21-41-45_faceshq_transformer/
```

--------------------------------

### Create Training Data List

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Generate a text file listing all JPG images in the training folder.

```bash
find $(pwd)/your_folder -name "*.jpg" > train.txt
```

--------------------------------

### Extracting Segmentation Masks Script

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

This script is an example for generating segmentation masks for the S-FLCKR dataset using DeepLab v2.

```python
scripts/extract_segmentation.py
```

--------------------------------

### Sample from FFHQ Model (Fast)

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Use this command to sample from the FFHQ model. The default parameters generate a large number of samples. Experiment with top-k and top-p for varied results.

```bash
python scripts/sample_fast.py -r logs/2021-04-23T18-19-01_ffhq_transformer/
```

--------------------------------

### Sample from ImageNet Model (Fast)

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Run this command to generate samples from the ImageNet model. Adjust parameters like -n (number of samples), -k (top-k), -t (temperature), and --batch_size as needed.

```bash
python scripts/sample_fast.py -r logs/2021-04-03T19-39-50_cin_transformer/ -n 50 -k 600 -t 1.0 -p 0.92 --batch_size 25
```

--------------------------------

### Sample from CelebA-HQ Model (Fast)

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Run this command to sample from the CelebA-HQ model. Similar to FFHQ, varying sampling parameters can yield different results.

```bash
python scripts/sample_fast.py -r logs/2021-04-23T18-11-19_celebahq_transformer/
```

--------------------------------

### KITTI Prompt Configuration

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Specifies classes and templates for the KITTI dataset. Suitable for car-related image descriptions.

```bash
classes = [
    'a photo i took of a car on my left or right side.',
    'a photo i took with a car nearby.',
    'a photo i took with a car in the distance.',
    'a photo i took with no car.',
]

templates = [
    '{}',
]
```

--------------------------------

### STL10 Dataset Classes and Templates

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Defines the list of classes and text templates for the STL10 dataset. This setup is useful for generating prompts for image classification models.

```python
classes = [
    'airplane',
    'bird',
    'car',
    'cat',
    'deer',
    'dog',
    'horse',
    'monkey',
    'ship',
    'truck',
]

templates = [
    'a photo of a {}.',
    'a photo of the {}.',
]
```

--------------------------------

### Generate Scene Samples with Pre-trained Model

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Generate scene images using a pre-trained model. Specify the output directory, path to the pre-trained model, and desired resolution.

```bash
python scripts/make_scene_samples.py --outdir=/some/outdir -r /path/to/pretrained/model --resolution=512,512
```

--------------------------------

### Run Streamlit Demo for Conditional Sampling

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Use this command to launch a Streamlit demo for conditional sampling. Replace the path with your downloaded model folder.

```bash
streamlit run scripts/sample_conditional.py -- -r logs/2020-11-09T13-31-51_sflckr/
```

--------------------------------

### Train Scene Image Transformer Model (Open Images)

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Run the training script for Open Images dataset. Ensure 'ckpt_path' and 'data_path' are correctly configured in the YAML file.

```bash
python main.py --base configs/open_images_scene_images_transformer.yaml -t True --gpus 0,
```

--------------------------------

### Creating ADE20k Symlink

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Create a symbolic link named 'ade20k_root' in the 'data' directory pointing to the contents of ADEChallengeData2016.zip.

```bash
ln -s /path/to/ADEChallengeData2016.zip data/ade20k_root
```

--------------------------------

### Run Streamlit Demo for D-RIN with Validation Set

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

To run the D-RIN demo on the complete validation set, first follow the ImageNet data preparation steps, then execute this command.

```bash
streamlit run scripts/sample_conditional.py -- -r logs/2020-11-20T12-54-32_drin_transformer/
```

--------------------------------

### Train Taming Transformers

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Run the main training script with a custom configuration. Use `--gpus 0,1` for multi-GPU training or `--gpus 0,` for single-GPU training.

```bash
python main.py --base configs/custom_vqgan.yaml -t True --gpus 0,1
```

--------------------------------

### Creating FFHQ Symlink

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Create a symbolic link named 'ffhq' in the 'data' directory pointing to the FFHQ images1024x1024 folder.

```bash
ln -s /path/to/images1024x1024 data/ffhq
```

--------------------------------

### DescribableTextures Classes and Templates

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Lists the texture classes and prompt templates for the DescribableTextures dataset. Use these to generate diverse prompts for CLIP.

```bash
classes = [
    'banded',
    'blotchy',
    'braided',
    'bubbly',
    'bumpy',
    'chequered',
    'cobwebbed',
    'cracked',
    'crosshatched',
    'crystalline',
    'dotted',
    'fibrous',
    'flecked',
    'freckled',
    'frilly',
    'gauzy',
    'grid',
    'grooved',
    'honeycombed',
    'interlaced',
    'knitted',
    'lacelike',
    'lined',
    'marbled',
    'matted',
    'meshed',
    'paisley',
    'perforated',
    'pitted',
    'pleated',
    'polka-dotted',
    'porous',
    'potholed',
    'scaly',
    'smeared',
    'spiralled',
    'sprinkled',
    'stained',
    'stratified',
    'striped',
    'studded',
    'swirly',
    'veined',
    'waffled',
    'woven',
    'wrinkled',
    'zigzagged',
]

templates = [
    'a photo of a {} texture.',
    'a photo of a {} pattern.',
    'a photo of a {} thing.',
    'a photo of a {} object.',
    'a photo of the {} texture.',
    'a photo of the {} pattern.',
    'a photo of the {} thing.',
    'a photo of the {} object.',
]
```

--------------------------------

### OxfordPets Dataset Classes and Templates

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Lists the pet breeds and a template for generating prompts for the OxfordPets dataset.

```bash
classes = [
    'Abyssinian',
    'Bengal',
    'Birman',
    'Bombay',
    'British Shorthair',
    'Egyptian Mau',
    'Maine Coon',
    'Persian',
    'Ragdoll',
    'Russian Blue',
    'Siamese',
    'Sphynx',
    'american bulldog',
    'american pit bull terrier',
    'basset hound',
    'beagle',
    'boxer',
    'chihuahua',
    'english cocker spaniel',
    'english setter',
    'german shorthaired',
    'great pyrenees',
    'havanese',
    'japanese chin',
    'keeshond',
    'leonberger',
    'miniature pinscher',
    'newfoundland',
    'pomeranian',
    'pug',
    'saint bernard',
    'samoyed',
    'scottish terrier',
    'shiba inu',
    'staffordshire bull terrier',
    'wheaten terrier',
    'yorkshire terrier',
]

templates = [
    'a photo of a {}, a type of pet.',
]
```

--------------------------------

### HatefulMemes Prompt Configuration

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Defines classes and templates for the HatefulMemes dataset. Use this for meme classification tasks.

```bash
classes = [
    'meme',
    'hatespeech meme',
]

templates = [
    'a {}.',
]
```

--------------------------------

### UCF101 Prompt Templates

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

A Python list of string templates for generating descriptive prompts for actions in the UCF101 dataset. These templates use placeholders for action class names.

```python
templates = [
    'a photo of a person {}.',
    'a video of a person {}.',
    'a example of a person {}.',
    'a demonstration of a person {}.',
    'a photo of the person {}.',
    'a video of the person {}.',
    'a example of the person {}.',
    'a demonstration of the person {}.',
    'a photo of a person using {}.',
    'a video of a person using {}.',
    'a example of a person using {}.',
    'a demonstration of a person using {}.',
    'a photo of the person using {}.',
    'a video of the person using {}.',
    'a example of the person using {}.',
    'a demonstration of the person using {}.',
    'a photo of a person doing {}.',
    'a video of a person doing {}.',
    'a example of a person doing {}.',
    'a demonstration of a person doing {}.',
    'a photo of the person doing {}.',
    'a video of the person doing {}.',
    'a example of the person doing {}.',
    'a demonstration of the person doing {}.',
    'a photo of a person during {}.',
    'a video of a person during {}.',
    'a example of a person during {}.',
    'a demonstration of a person during {}.',
    'a photo of the person during {}.',
    'a video of the person during {}.',
    'a example of the person during {}.',
    'a demonstration of the person during {}.',
    'a photo of a person performing {}.',
    'a video of a person performing {}.',
    'a example of a person performing {}.',
    'a demonstration of a person performing {}.',
    'a photo of the person performing {}.',
    'a video of the person performing {}.',
    'a example of the person performing {}.',
    'a demonstration of the person performing {}.',
    'a photo of a person practicing {}.',
    'a video of a person practicing {}.',
    'a example of a person practicing {}.',
    'a demonstration of a person practicing {}.',
    'a photo of the person practicing {}.',
    'a video of the person practicing {}.',
    'a example of the person practicing {}.',
    'a demonstration of the person practicing {}.',
]
```

--------------------------------

### FacialEmotionRecognition2013 Prompt Templates

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Defines lists of emotions and corresponding prompt templates for the FacialEmotionRecognition2013 dataset. Use these to generate descriptive text for images.

```bash
classes = [
    ['angry'],
    ['disgusted'],
    ['fearful'],
    ['happy', 'smiling'],
    ['sad', 'depressed'],
    ['surprised', 'shocked', 'spooked'],
    ['neutral', 'bored'],
]

templates = [
    'a photo of a {} looking face.',
    'a photo of a face showing the emotion: {}.',
    'a photo of a face looking {}.',
    'a face that looks {}.',
    'they look {}.',
    'look at how {} they are.',
]
```

--------------------------------

### Train ImageNet VQGAN

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Command to initiate training of a VQGAN model on the ImageNet dataset. Pre-trained models can be downloaded and placed in the 'logs' directory.

```bash
python main.py --base configs/imagenet_vqgan.yaml -t True --gpus 0,
```

--------------------------------

### Creating CelebA-HQ Symlink

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Create a symbolic link named 'celebahq' in the 'data' directory pointing to the CelebA-HQ .npy files.

```bash
ln -s /path/to/celebahq data/celebahq
```

--------------------------------

### EuroSAT Classes and Templates

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Lists the land cover classes and prompt templates for the EuroSAT dataset. Useful for generating prompts for satellite imagery analysis.

```bash
classes = [
    'forest',
    'permanent crop land',
    'residential buildings or homes or apartments',
    'river',
    'pasture land',
    'lake or sea',
    'brushland or shrubland',
    'annual crop land',
    'industrial buildings or commercial buildings',
    'highway or road',
]

templates = [
    'a centered satellite photo of {}.',
    'a centered satellite photo of a {}.',
    'a centered satellite photo of the {}.',
]
```

--------------------------------

### Train FacesHQ VQGAN

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Use this command to train a VQGAN model on the FacesHQ dataset. Ensure you have the correct configuration file.

```bash
python main.py --base configs/faceshq_vqgan.yaml -t True --gpus 0,
```

--------------------------------

### Train Scene Image Transformer Model (COCO)

Source: https://github.com/biomap-research/morphdiff/blob/main/src/taming-transformers/README.md

Run the training script for COCO dataset. Ensure 'ckpt_path' and 'data_path' are correctly configured in the YAML file.

```bash
python main.py --base configs/coco_scene_images_transformer.yaml -t True --gpus 0,
```

--------------------------------

### PascalVOC2007 Dataset Classes and Templates

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Specifies the object classes and a template for generating prompts for the PascalVOC2007 dataset.

```bash
classes = [
    'aeroplane',
    'bicycle',
    'bird',
    'boat',
    'bottle',
    'bus',
    'car',
    'cat',
    'chair',
    'cow',
    'dog',
    'horse',
    'motorbike',
    'person',
    'sheep',
    'sofa',
    'diningtable',
    'pottedplant',
    'train',
    'tvmonitor',
]

templates = [
    'a photo of a {}.',
]
```

--------------------------------

### Image Generation Templates

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

These templates are used to construct prompts for image generation. They incorporate a placeholder '{}' which will be replaced by a scene category from the SUN397 dataset.

```python
templates = [
    'a photo of a {}.',
    'a photo of the {}.',
]
```

--------------------------------

### Download and Extract Rendered SST2 Dataset

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/rendered-sst2.md

Use these bash commands to download the 131MB archive containing the Rendered SST2 images and extract them into a subdirectory named 'rendered-sst2'.

```bash
wget https://openaipublic.azureedge.net/clip/data/rendered-sst2.tgz
tar zxvf rendered-sst2.tgz
```

--------------------------------

### Flowers102 Prompt Templates

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Lists flower names and a prompt template for the Flowers102 dataset. This helps in creating text descriptions for flower images.

```bash
classes = [
    'pink primrose',
    'hard-leaved pocket orchid',
    'canterbury bells',
    'sweet pea',
    'english marigold',
    'tiger lily',
    'moon orchid',
    'bird of paradise',
    'monkshood',
    'globe thistle',
    'snapdragon',
    "colt's foot",
    'king protea',
    'spear thistle',
    'yellow iris',
    'globe flower',
    'purple coneflower',
    'peruvian lily',
    'balloon flower',
    'giant white arum lily',
    'fire lily',
    'pincushion flower',
    'fritillary',
    'red ginger',
    'grape hyacinth',
    'corn poppy',
    'prince of wales feathers',
    'stemless gentian',
    'artichoke',
    'sweet william',
    'carnation',
    'garden phlox',
    'love in the mist',
    'mexican aster',
    'alpine sea holly',
    'ruby-lipped cattleya',
    'cape flower',
    'great masterwort',
    'siam tulip',
    'lenten rose',
    'barbeton daisy',
    'daffodil',
    'sword lily',
    'poinsettia',
    'bolero deep blue',
    'wallflower',
    'marigold',
    'buttercup',
    'oxeye daisy',
    'common dandelion',
    'petunia',
    'wild pansy',
    'primula',
    'sunflower',
    'pelargonium',
    'bishop of llandaff',
    'gaura',
    'geranium',
    'orange dahlia',
    'pink and yellow dahlia',
    'cautleya spicata',
    'japanese anemone',
    'black-eyed susan',
    'silverbush',
    'californian poppy',
    'osteospermum',
    'spring crocus',
    'bearded iris',
    'windflower',
    'tree poppy',
    'gazania',
    'azalea',
    'water lily',
    'rose',
    'thorn apple',
    'morning glory',
    'passion flower',
    'lotus',
    'toad lily',
    'anthurium',
    'frangipani',
    'clematis',
    'hibiscus',
    'columbine',
    'desert-rose',
    'tree mallow',
    'magnolia',
    'cyclamen',
    'watercress',
    'canna lily',
    'hippeastrum',
    'bee balm',
    'air plant',
    'foxglove',
    'bougainvillea',
    'camellia',
    'mallow',
    'mexican petunia',
    'bromelia',
    'blanket flower',
    'trumpet creeper',
    'blackberry lily',
]

templates = [
    'a photo of a {}, a type of flower.',
]
```

--------------------------------

### PatchCamelyon Dataset Classes and Templates

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Defines the tissue classes and a template for generating prompts for the PatchCamelyon dataset.

```bash
classes = [
    'lymph node',
    'lymph node containing metastatic tumor tissue',
]

templates = [
    'this is a photo of {}',
]
```

--------------------------------

### Caltech101 Class and Template Prompts

Source: https://github.com/biomap-research/morphdiff/blob/main/src/clip/data/prompts.md

Provides the class names and a comprehensive set of text templates for the Caltech101 dataset. These are used for generating diverse prompts for various object categories.

```bash
classes = [
    'background',
    'off-center face',
    'centered face',
    'leopard',
    'motorbike',
    'accordion',
    'airplane',
    'anchor',
    'ant',
    'barrel',
    'bass',
    'beaver',
    'binocular',
    'bonsai',
    'brain',
    'brontosaurus',
    'buddha',
    'butterfly',
    'camera',
    'cannon',
    'side of a car',
    'ceiling fan',
    'cellphone',
    'chair',
    'chandelier',
    'body of a cougar cat',
    'face of a cougar cat',
    'crab',
    'crayfish',
    'crocodile',
    'head of a  crocodile',
    'cup',
    'dalmatian',
    'dollar bill',
    'dolphin',
    'dragonfly',
    'electric guitar',
    'elephant',
    'emu',
    'euphonium',
    'ewer',
    'ferry',
    'flamingo',
    'head of a flamingo',
    'garfield',
    'gerenuk',
    'gramophone',
    'grand piano',
    'hawksbill',
    'headphone',
    'hedgehog',
    'helicopter',
    'ibis',
    'inline skate',
    'joshua tree',
    'kangaroo',
    'ketch',
    'lamp',
    'laptop',
    'llama',
    'lobster',
    'lotus',
    'mandolin',
    'mayfly',
    'menorah',
    'metronome',
    'minaret',
    'nautilus',
    'octopus',
    'okapi',
    'pagoda',
    'panda',
    'pigeon',
    'pizza',
    'platypus',
    'pyramid',
    'revolver',
    'rhino',
    'rooster',
    'saxophone',
    'schooner',
    'scissors',
    'scorpion',
    'sea horse',
    'snoopy (cartoon beagle)',
    'soccer ball',
    'stapler',
    'starfish',
    'stegosaurus',
    'stop sign',
    'strawberry',
    'sunflower',
    'tick',
    'trilobite',
    'umbrella',
    'watch',
    'water lilly',
    'wheelchair',
    'wild cat',
    'windsor chair',
    'wrench',
    'yin and yang symbol',
]

templates = [
    'a photo of a {}.',
    'a painting of a {}.',
    'a plastic {}.',
    'a sculpture of a {}.',
    'a sketch of a {}.',
    'a tattoo of a {}.',
    'a toy {}.',
    'a rendition of a {}.',
    'a embroidered {}.',
    'a cartoon {}.',
    'a {} in a video game.',
    'a plushie {}.',
    'a origami {}.',
    'art of a {}.',
    'graffiti of a {}.',
    'a drawing of a {}.',
    'a doodle of a {}.',
    'a photo of the {}.',
    'a painting of the {}.',
    'the plastic {}.',
    'a sculpture of the {}.',
    'a sketch of the {}.',
    'a tattoo of the {}.',
    'the toy {}.',
    'a rendition of the {}.',
    'the embroidered {}.',
    'the cartoon {}.',
    'the {} in a video game.',
    'the plushie {}.',
    'the origami {}.',
    'art of the {}.',
    'graffiti of the {}.',
    'a drawing of the {}.',
    'a doodle of the {}.',
]
```