app.py

import gradio as gr
import torch
from diffusers import StableDiffusionPipeline, DPMSolverMultistepScheduler

import numpy as np
import os
import cv2
from PIL import Image, ImageDraw
import insightface
from insightface.app import FaceAnalysis

# Diffusion
model_base = "runwayml/stable-diffusion-v1-5"

pipe = StableDiffusionPipeline.from_pretrained(model_base, torch_dtype=torch.float16, use_safetensors=True, safety_checker=None,)
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)

lora_model_path = "./loralucy3/checkpoint-95000"
pipe.unet.load_attn_procs(lora_model_path)
pipe.to("cuda")


# Insightface model
app = FaceAnalysis(name='buffalo_l')
app.prepare(ctx_id=0, det_size=(640, 640))


def face_swap(src_img, dest_img):
    src_img = Image.open('./images/' + src_img + '.JPG')
    
    # Convert to RGB
    src_img = src_img.convert(mode='RGB')
    dest_img = dest_img.convert(mode='RGB')

    # Convert to array
    src_img_arr = np.asarray(src_img)
    dest_img_arr = np.asarray(dest_img)

    # Face detection
    src_faces = app.get(src_img_arr)
    dest_faces = app.get(dest_img_arr)

    # Initialize swapper
    swapper = insightface.model_zoo.get_model('inswapper_128.onnx', download=False, download_zip=False)

    # Swap face
    res = dest_img_arr.copy()
    for face in dest_faces:
    	res = swapper.get(res, face, src_faces[0], paste_back=True)

    # Convert to PIL image
    final_image = Image.fromarray(np.uint8(res)).convert('RGB')

    return final_image 


def greet(description,color,features,occasion,type_,face):

    # Parse input
    prompt = 'white background '
    description = 'description:' + description.replace(' ', '-')
    color = ' color:' + ','.join(color)
    features = ' features:' + ','.join(features)
    occasion = ' occasion:' + ','.join(occasion)
    type_ = ' type:' + ','.join(type_)
   
    prompt += description + color + features +  occasion + type_

    print('prompt:',prompt)
    image = pipe(
        prompt,
        negative_prompt='deformed face,bad anatomy',
        width=312,
        height=512,
        num_inference_steps=100,
        guidance_scale=7.5,
        cross_attention_kwargs={"scale": 1.0}
        ).images[0]

    if(face != 'Normal'):
        image = face_swap(face, image)
    
    return image

iface = gr.Interface(fn=greet, 
                     inputs=[gr.Textbox(label='Description'),
                            gr.Dropdown(label='Color',choices=['Beige','Black','Blue','Brown','Green','Grey','Orange','Pink','Purple','Red','White','Yellow'],multiselect=True),
                            gr.Dropdown(label='Features',choices=['3/4-sleeve','Babydoll','Closed-Back','Corset','Crochet','Cutouts','Draped','Floral','Gloves','Halter','Lace','Long','Long-Sleeve','Midi','No-Slit','Off-The-Shoulder','One-Shoulder','Open-Back','Pockets','Print','Puff-Sleeve','Ruched','Satin','Sequins','Shimmer','Short','Short-Sleeve','Side-Slit','Square-Neck','Strapless','Sweetheart-Neck','Tight','V-Neck','Velvet','Wrap'],multiselect=True),
                            gr.Dropdown(label='Occasion',choices=['Homecoming','Casual','Wedding-Guest','Festival','Sorority','Day','Vacation','Summer','Pool-Party','Birthday','Date-Night','Party','Holiday','Winter-Formal','Valentines-Day','Prom','Graduation'],multiselect=True),
                            gr.Dropdown(label='Type',choices=['Mini-Dresses','Midi-Dresses','Maxi-Dresses','Two-Piece-Sets','Rompers','Jeans','Jumpsuits','Pants','Tops','Jumpers/Cardigans','Skirts','Shorts','Bodysuits','Swimwear'],multiselect=True),
                            gr.Dropdown(label='Face',choices=['Normal','Cat','Lisa','Mila'], value='Normal'),
                            ], 
                     outputs=gr.Image(type="pil", label="Final Image", width=312, height=512))
iface.launch()