app.py

import jax
import jax.numpy as jnp
from flax import jax_utils
from flax.training.common_utils import shard
from PIL import Image
from argparse import Namespace
import gradio as gr
import copy # added
import numpy as np
import mediapipe as mp
from mediapipe import solutions
from mediapipe.framework.formats import landmark_pb2
from mediapipe.tasks import python
from mediapipe.tasks.python import vision
import cv2
import psutil
from gpuinfo import GPUInfo
import time
import gc
import torch

from diffusers import (
    FlaxControlNetModel,
    FlaxStableDiffusionControlNetPipeline,
)
right_style_lm = copy.deepcopy(solutions.drawing_styles.get_default_hand_landmarks_style())
left_style_lm = copy.deepcopy(solutions.drawing_styles.get_default_hand_landmarks_style())
right_style_lm[0].color=(251, 206, 177)
left_style_lm[0].color=(255, 255, 225)

def draw_landmarks_on_image(rgb_image, detection_result, overlap=False, hand_encoding=False):
    hand_landmarks_list = detection_result.hand_landmarks
    handedness_list = detection_result.handedness
    if overlap:
        annotated_image = np.copy(rgb_image)
    else:
        annotated_image = np.zeros_like(rgb_image)

    # Loop through the detected hands to visualize.
    for idx in range(len(hand_landmarks_list)):
        hand_landmarks = hand_landmarks_list[idx]
        handedness = handedness_list[idx]
        # Draw the hand landmarks.
        hand_landmarks_proto = landmark_pb2.NormalizedLandmarkList()
        hand_landmarks_proto.landmark.extend([
                landmark_pb2.NormalizedLandmark(x=landmark.x, y=landmark.y, z=landmark.z) for landmark in hand_landmarks
        ])
        if hand_encoding:
            if handedness[0].category_name == "Left":
                solutions.drawing_utils.draw_landmarks(
                        annotated_image,
                        hand_landmarks_proto,
                        solutions.hands.HAND_CONNECTIONS,
                        left_style_lm,
                        solutions.drawing_styles.get_default_hand_connections_style())
            if handedness[0].category_name == "Right":
                solutions.drawing_utils.draw_landmarks(
                        annotated_image,
                        hand_landmarks_proto,
                        solutions.hands.HAND_CONNECTIONS,
                        right_style_lm,
                        solutions.drawing_styles.get_default_hand_connections_style())
        else:
            solutions.drawing_utils.draw_landmarks(
                    annotated_image,
                    hand_landmarks_proto,
                    solutions.hands.HAND_CONNECTIONS,
                    solutions.drawing_styles.get_default_hand_landmarks_style(),
                    solutions.drawing_styles.get_default_hand_connections_style())

    return annotated_image

def generate_annotation(img, overlap=False, hand_encoding=False):
    """img(input): numpy array
       annotated_image(output): numpy array
    """
    # STEP 2: Create an HandLandmarker object.
    base_options = python.BaseOptions(model_asset_path='hand_landmarker.task')
    options = vision.HandLandmarkerOptions(base_options=base_options,
                                        num_hands=2)
    detector = vision.HandLandmarker.create_from_options(options)

    # STEP 3: Load the input image.
    image = mp.Image(
        image_format=mp.ImageFormat.SRGB, data=img)

    # STEP 4: Detect hand landmarks from the input image.
    detection_result = detector.detect(image)

    # STEP 5: Process the classification result. In this case, visualize it.
    annotated_image = draw_landmarks_on_image(image.numpy_view(), detection_result, overlap=overlap, hand_encoding=hand_encoding)
    return annotated_image



std_args = Namespace(
        pretrained_model_name_or_path="runwayml/stable-diffusion-v1-5",
        revision="non-ema",
        from_pt=True,
        controlnet_model_name_or_path="Vincent-luo/controlnet-hands",
        controlnet_revision=None,
        controlnet_from_pt=False,
    )
enc_args = Namespace(
        pretrained_model_name_or_path="runwayml/stable-diffusion-v1-5",
        revision="non-ema",
        from_pt=True,
        controlnet_model_name_or_path="MakiPan/controlnet-encoded-hands-130k",
        controlnet_revision=None,
        controlnet_from_pt=False,
    )

std_controlnet, std_controlnet_params = FlaxControlNetModel.from_pretrained(
    std_args.controlnet_model_name_or_path,
    revision=std_args.controlnet_revision,
    from_pt=std_args.controlnet_from_pt,
    dtype=jnp.float32, # jnp.bfloat16
)
enc_controlnet, enc_controlnet_params = FlaxControlNetModel.from_pretrained(
    enc_args.controlnet_model_name_or_path,
    revision=enc_args.controlnet_revision,
    from_pt=enc_args.controlnet_from_pt,
    dtype=jnp.float32, # jnp.bfloat16
)



std_pipeline, std_pipeline_params = FlaxStableDiffusionControlNetPipeline.from_pretrained(
    std_args.pretrained_model_name_or_path,
    # tokenizer=tokenizer,
    controlnet=std_controlnet,
    safety_checker=None,
    dtype=jnp.float32, # jnp.bfloat16
    revision=std_args.revision,
    from_pt=std_args.from_pt,
)
enc_pipeline, enc_pipeline_params = FlaxStableDiffusionControlNetPipeline.from_pretrained(
    enc_args.pretrained_model_name_or_path,
    # tokenizer=tokenizer,
    controlnet=enc_controlnet,
    safety_checker=None,
    dtype=jnp.float32, # jnp.bfloat16
    revision=enc_args.revision,
    from_pt=enc_args.from_pt,
)


std_pipeline_params["controlnet"] = std_controlnet_params
std_pipeline_params = jax_utils.replicate(std_pipeline_params)

enc_pipeline_params["controlnet"] = enc_controlnet_params
enc_pipeline_params = jax_utils.replicate(enc_pipeline_params)

rng = jax.random.PRNGKey(0)
num_samples = jax.device_count()
prng_seed = jax.random.split(rng, jax.device_count())
memory = psutil.virtual_memory()

def infer(prompt, negative_prompt, image, model_type="Standard"):
    time_start = time.time()
    prompts = num_samples * [prompt]
    if model_type=="Standard":
        prompt_ids = std_pipeline.prepare_text_inputs(prompts)
    elif model_type=="Hand Encoding":
        prompt_ids = enc_pipeline.prepare_text_inputs(prompts)
    else:
        pass
    prompt_ids = shard(prompt_ids)

    if model_type=="Standard":
        annotated_image = generate_annotation(image, overlap=False, hand_encoding=False)
        overlap_image = generate_annotation(image, overlap=True, hand_encoding=False)
    elif model_type=="Hand Encoding":
        annotated_image = generate_annotation(image, overlap=False, hand_encoding=True)
        overlap_image = generate_annotation(image, overlap=True, hand_encoding=True)

    else:
        pass
    validation_image = Image.fromarray(annotated_image).convert("RGB")

    if model_type=="Standard":
        processed_image = std_pipeline.prepare_image_inputs(num_samples * [validation_image])
        processed_image = shard(processed_image)

        negative_prompt_ids = std_pipeline.prepare_text_inputs([negative_prompt] * num_samples)
        negative_prompt_ids = shard(negative_prompt_ids)

        images = std_pipeline(
            prompt_ids=prompt_ids,
            image=processed_image,
            params=std_pipeline_params,
            prng_seed=prng_seed,
            num_inference_steps=50,
            neg_prompt_ids=negative_prompt_ids,
            jit=True,
        ).images
    elif model_type=="Hand Encoding":
        processed_image = enc_pipeline.prepare_image_inputs(num_samples * [validation_image])
        processed_image = shard(processed_image)

        negative_prompt_ids = enc_pipeline.prepare_text_inputs([negative_prompt] * num_samples)
        negative_prompt_ids = shard(negative_prompt_ids)

        images = enc_pipeline(
            prompt_ids=prompt_ids,
            image=processed_image,
            params=enc_pipeline_params,
            prng_seed=prng_seed,
            num_inference_steps=50,
            neg_prompt_ids=negative_prompt_ids,
            jit=True,
        ).images

    else:
        pass
    images = images.reshape((images.shape[0] * images.shape[1],) + images.shape[-3:])

    results = [i for i in images]
    
    # running info
    time_end = time.time()
    time_diff = time_end - time_start
    gc.collect()
    torch.cuda.empty_cache()
    memory = psutil.virtual_memory()
    gpu_utilization, gpu_memory = GPUInfo.gpu_usage()
    gpu_utilization = gpu_utilization[0] if len(gpu_utilization) > 0 else 0
    gpu_memory = gpu_memory[0] if len(gpu_memory) > 0 else 0
    system_info = f"""
    *Memory: {memory.total / (1024 * 1024 * 1024):.2f}GB, used: {memory.percent}%, available: {memory.available / (1024 * 1024 * 1024):.2f}GB.* 
    *Processing time: {time_diff:.5} seconds.*
    *GPU Utilization: {gpu_utilization}%, GPU Memory: {gpu_memory}MiB.*
    """
    return [overlap_image, annotated_image] + results, system_info 


with gr.Blocks(theme='gradio/soft') as demo:
    gr.Markdown("## Stable Diffusion with Hand Control")
    gr.Markdown("This model is a ControlNet model using MediaPipe hand landmarks for control.")
    with gr.Box():
        gr.Markdown("""<h2><b>Summary 📋</b></h2>""")
        with gr.Accordion("Detail information", open=False):
            gr.Markdown("""
        As Stable diffusion and other diffusion models are notoriously poor at generating realistic hands for our project we decided to train a ControlNet model using MediaPipes landmarks in order to generate more realistic hands avoiding common issues such as unrealistic positions and irregular digits.
        <br>
        We opted to use the [HAnd Gesture Recognition Image Dataset](https://github.com/hukenovs/hagrid) (HaGRID) and [MediaPipe's Hand Landmarker](https://developers.google.com/mediapipe/solutions/vision/hand_landmarker) to train a control net that could potentially be used independently or as an in-painting tool.
        To preprocess the data there were three options we considered:
        <ul>
            <li>The first was to use Mediapipes built-in draw landmarks function. This was an obvious first choice however we noticed with low training steps that the model couldn't easily distinguish handedness and would often generate the wrong hand for the conditioning image.</li>
        <center>
        <table><tr>
        <td> 
          <p align="center" style="padding: 10px">
            <img alt="Forwarding" src="https://datasets-server.huggingface.co/assets/MakiPan/hagrid250k-blip2/--/MakiPan--hagrid250k-blip2/train/29/image/image.jpg" width="200">
            <br>
            <em style="color: grey">Original Image</em>
          </p> 
        </td>
        <td> 
          <p align="center">
            <img alt="Routing" src="https://datasets-server.huggingface.co/assets/MakiPan/hagrid250k-blip2/--/MakiPan--hagrid250k-blip2/train/29/conditioning_image/image.jpg" width="200">
            <br>
            <em style="color: grey">Conditioning Image</em>
          </p> 
        </td>
        </tr></table>
        </center>
            <li>To counter this issue we changed the palm landmark colors with the intention to keep the color similar in order to learn that they provide similar information, but different to make the model know which hands were left or right.</li>   
        <center>
        <table><tr>
        <td> 
          <p align="center" style="padding: 10px">
            <img alt="Forwarding" src="https://datasets-server.huggingface.co/assets/MakiPan/hagrid-hand-enc-250k/--/MakiPan--hagrid-hand-enc-250k/train/96/image/image.jpg" width="200">
            <br>
            <em style="color: grey">Original Image</em>
          </p> 
        </td>
        <td> 
          <p align="center">
            <img alt="Routing" src="https://datasets-server.huggingface.co/assets/MakiPan/hagrid-hand-enc-250k/--/MakiPan--hagrid-hand-enc-250k/train/96/conditioning_image/image.jpg" width="200">
            <br>
            <em style="color: grey">Conditioning Image</em>
          </p> 
        </td>
        </tr></table>
        </center>
            <li>The last option was to use <a href="https://ai.googleblog.com/2020/12/mediapipe-holistic-simultaneous-face.html">MediaPipe Holistic</a> to provide pose face and hand landmarks to the ControlNet. This method was promising in theory, however, the HaGRID dataset was not suitable for this method as the Holistic model performs poorly with partial body and obscurely cropped images.</li>
        </ul>
        We anecdotally determined that when trained at lower steps the encoded hand model performed better than the standard MediaPipe model due to implied handedness. We theorize that with a larger dataset of more full-body hand and pose classifications, Holistic landmarks will provide the best images in the future however for the moment the hand-encoded model performs best.
        """)
    
    # Information links
    with gr.Box():
        gr.Markdown("""<h2><b>Links 🔗</b></h2>""")
        with gr.Accordion("Models 🚀", open=False):
            gr.Markdown("""
            <h4><a href="https://huggingface.co./Vincent-luo/controlnet-hands">Standard Model</a></h4>
            <h4> <a href="https://huggingface.co./MakiPan/controlnet-encoded-hands-130k/">Model using Hand Encoding</a></h4>
            """)

        with gr.Accordion("Datasets 💾", open=False):
            gr.Markdown("""
            <h4> <a href="https://huggingface.co./datasets/MakiPan/hagrid250k-blip2">Dataset for Standard Model</a></h4>
            <h4> <a href="https://huggingface.co./datasets/MakiPan/hagrid-hand-enc-250k">Dataset for Hand Encoding Model</a></h4>
            """)

        with gr.Accordion("Preprocessing Scripts 📑", open=False):
            gr.Markdown("""
            <h4> <a href="https://github.com/Maki-DS/Jax-Controlnet-hand-training/blob/main/normal-preprocessing.py">Standard Data Preprocessing Script</a></h4>
            <h4> <a href="https://github.com/Maki-DS/Jax-Controlnet-hand-training/blob/main/Hand-encoded-preprocessing.py">Hand Encoding Data Preprocessing Script</a></h4></center>
            """)
            
    # How to use model
    with gr.Box():
        gr.Markdown("""<h2><b>How to use ⌛️</b></h2>""")
        with gr.Accordion("Generate image with ControlnetHand", open=True):
            gr.Markdown("""
            - Step 1. Select preprocessing method (Standard or Hand encoding)
            - Step 2. Describe the image you want to create along with the hand details of the uploaded or captured image
            - Step 3. Provide a negative prompt that helps the model not to create redundant details
            - Step 4. Upload or capture by webcam a clear image of hands that are prominently visible in the foreground
            - Step 5. Submit and enjoy
            """)
            
    # Model input parameters
    model_type = gr.Radio(["Standard", "Hand Encoding"], value="Standard", label="Model preprocessing", info="We developed two models, one with standard MediaPipe landmarks, and one with different (but similar) coloring on palm landmarks to distinguish left and right")
    
    with gr.Row():
        with gr.Column():
            prompt_input = gr.Textbox(label="Prompt")
            negative_prompt = gr.Textbox(label="Negative Prompt")
            with gr.Box():
                with gr.Tab("Upload Image"):
                    upload_image = gr.Image(label="Upload Image", source="upload")
                with gr.Tab("Webcam"):
                    webcam_image = gr.Image(label="Webcam", source="webcam")
            # output_image = gr.Gallery(label='Output Image', show_label=False, elem_id="gallery").style(grid=3, height='auto')
            submit_btn = gr.Button(value = "Submit")
            # inputs = [prompt_input, negative_prompt, input_image]
            # submit_btn.click(fn=infer, inputs=inputs, outputs=[output_image])
            system_info = gr.Markdown(f"*Memory: {memory.total / (1024 * 1024 * 1024):.2f}GB, used: {memory.percent}%, available: {memory.available / (1024 * 1024 * 1024):.2f}GB*")
        with gr.Column():
            output_image = gr.Gallery(label='Output Image', show_label=False, elem_id="gallery").style(grid=2, height='auto')
    
    gr.Examples(
        examples=[
            [
               "a woman is making an ok sign in front of a painting",
               "longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality",
               "example.png"
            ],
            [
               "a man with his hands up in the air making a rock sign",
               "longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality",
               "example1.png"
            ],
            [
               "a man is making a thumbs up gesture",
               "longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality",
               "example2.png"
            ],
            [
               "a woman is holding up her hand in front of a window",
               "longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality",
               "example3.png"
            ],
            [
               "a man with his finger on his lips",
               "longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality",
               "example4.png"
            ],
        ],
        inputs=[prompt_input, negative_prompt, upload_image, model_type],
        outputs=[output_image, system_info],
        fn=infer,
        cache_examples=True,
    )
    # check source of image
    if upload_image and webcam_image is None:
        input_image = upload_image
    else:
        input_image = webcam_image
        
    inputs = [prompt_input, negative_prompt, input_image, model_type]
    submit_btn.click(fn=infer, inputs=inputs, outputs=[output_image, system_info])

demo.launch()