app.py

import gradio as gr
import torch.nn as nn
import torch
from src.core.config import BaseConfig
from src.core.yaml_utils import load_config, merge_config, create, merge_dict
import cv2
import torchvision.transforms as transforms
from PIL import Image, ImageDraw
from src.core import YAMLConfig
from pathlib import Path

transformer = transforms.Compose([    
    transforms.ToTensor(),                            
    #transforms.Resize([640,640]),

])


#def model(yaml_cfg) -> torch.nn.Module:
#    if 'model' in yaml_cfg:
#        merge_config(yaml_cfg)
#        model = create(yaml_cfg['model'])
#    return model
#
#
#
#def get_model(cfg_path='./configs/rtdetr/rtdetr_r101vd_6x_coco.yml',ckpt="./logs/checkpointcope12.pth"):
#  yaml_cfg = load_config(cfg_path)
#  merge_dict(yaml_cfg, {}) 
#  tmodel = model(yaml_cfg)
#  checkpoint = torch.load(ckpt, map_location='cpu') 
#  if 'ema' in checkpoint:
#      state = checkpoint['ema']['module']
#  else:
#      state = checkpoint['model']
#      
#  tmodel.load_state_dict(state)
#  return tmodel
  
class Model(nn.Module):
    def __init__(self, confg=None, ckpt="") -> None:
        super().__init__()
        self.cfg = YAMLConfig(confg, resume=ckpt)
        if ckpt:
            checkpoint = torch.load(ckpt, map_location='cpu') 
            if 'ema' in checkpoint:
                state = checkpoint['ema']['module']
            else:
                state = checkpoint['model']
        else:
            raise AttributeError('only support resume to load model.state_dict by now.')

        # NOTE load train mode state -> convert to deploy mode
        self.cfg.model.load_state_dict(state)

        self.model = self.cfg.model.deploy()
        self.postprocessor = self.cfg.postprocessor.deploy()
        # print(self.postprocessor.deploy_mode)
        
    def forward(self, images, orig_target_sizes):
        outputs = self.model(images)
        return self.postprocessor(outputs, orig_target_sizes)
  

model = Model(confg='./configs/rtdetr/rtdetr_r101vd_6x_coco.yml',ckpt="./checkpointcope12.pth")

#img = cv2.imread('./j.jpg',cv2.IMREAD_GRAYSCALE)
#img = Image.open('./a.jpg').convert('RGB').resize((640,640))


def detect(img,thr):
  #print(img)                    #ndarray
  img = Image.fromarray(img).resize((640,640))
  t_img = transformer(img).unsqueeze(0)#.unsqueeze(0)      #[1,1,640,640]
  size = torch.tensor([[t_img.shape[2], t_img.shape[3]]])
  #print(t_img.shape)
  labels, boxes, scores=model(t_img,size)
  #img_path = Path('./a.jpg')
  draw = ImageDraw.Draw(img)
  thrh = thr
  
  for i in range(t_img.shape[0]):
  
      scr = scores[i]
      lab = labels[i][scr > thrh]
      box = boxes[i][scr > thrh]
  
  
      for b in box:
          draw.rectangle(list(b), outline='red', )
          draw.text((b[0], b[1]), text=str(lab[i]), fill='blue', )
  
  #save_path = Path('./output') / img_path.name
  return img

interface = gr.Interface(fn=detect,inputs=["image",gr.Slider(label=“thr”, value=0.2, maximum=1, minimum=0)],outputs="image",title="rt-cope detect")
  
interface.launch()