import tqdm | |
#import fastCNN | |
import numpy as np | |
import gradio as gr | |
import os | |
#os.system("sudo apt-get install nvIDia-cuda-toolkit") | |
#os.system("/usr/local/bin/python -m pip install --upgrade pip") | |
#os.system("pip install argparse") | |
os.system("pip install opencv-python") | |
#import pydensecrf.densecrf as dcrf | |
from PIL import Image | |
#import torch | |
#import torch.nn.functional as F | |
#from torchvision import transforms | |
import numpy as np | |
import collections | |
import cv2 | |
#import argparse | |
device='cpu' | |
def test(gpu_id, net, img_list, group_size, img_size): | |
print('test') | |
#device=device | |
hl,wl=[_.shape[0] for _ in img_list],[_.shape[1] for _ in img_list] | |
img_transform = transforms.Compose([transforms.Resize((img_size, img_size)), transforms.ToTensor(), | |
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])]) | |
img_transform_gray = transforms.Compose([transforms.Resize((img_size, img_size)), transforms.ToTensor(), | |
transforms.Normalize(mean=[0.449], std=[0.226])]) | |
with torch.no_grad(): | |
group_img=torch.rand(5,3,224,224) | |
for i in range(5): | |
group_img[i]=img_transform(Image.fromarray(img_list[i])) | |
_,pred_mask=net(group_img*1) | |
pred_mask=(pred_mask.detach().squeeze()*255)#.numpy().astype(np.uint8) | |
#pred_mask=[F.interpolate(pred_mask[i].reshape(1,1,pred_mask[i].shape[-2],pred_mask[i].shape[-1]),size=(size,size),mode='bilinear').squeeze().numpy().astype(np.uint8) for i in range(5)] | |
img_resize=[((group_img[i]-group_img[i].min())/(group_img[i].max()-group_img[i].min())*255).permute(1,2,0).contiguous().numpy().astype(np.uint8) | |
for i in range(5)] | |
pred_mask=[crf_refine(img_resize[i],pred_mask[i].numpy().astype(np.uint8)) for i in range(5)] | |
#for i in range(5): | |
# print(img_list[i].shape,pred_mask[i].shape) | |
#pred_mask=[crf_refine(img_list[i],pred_mask[i]) for i in range(5)] | |
print(pred_mask[0].shape) | |
white=(torch.ones(2,pred_mask[0].shape[1],3)*255).long() | |
result = [torch.cat([torch.from_numpy(img_resize[i]),white,torch.from_numpy(pred_mask[i]).unsqueeze(2).repeat(1,1,3)],dim=0).numpy() for i in range(5)] | |
#w, h = 224,224#Image.open(image_list[i][j]).size | |
#result = result.resize((w, h), Image.BILINEAR) | |
#result.convert('L').save('0.png') | |
print('done') | |
return result | |
#img_lst=[(torch.rand(352,352,3)*255).numpy().astype(np.uint8) for i in range(5)] | |
outputpath1='img2.png' | |
outputpath2='img2.png' | |
outputpath3='img2.png' | |
def sepia(opt,img1): | |
#img_list=[img1,img2,img3,img4,img5] | |
#h_list,w_list=[_.shape[0] for _ in img_list],[_.shape[1] for _ in img_list] | |
#print(type(img1)) | |
#print(img1.shape) | |
#result_list=test(device,net,img_list,5,224) | |
#result_list=[result_list[i].resize((w_list[i], h_list[i]), Image.BILINEAR) for i in range(5)] | |
#img1,img2,img3,img4,img5=result_list#test('cpu',net,img_list,5,224) | |
#white=(torch.ones(img1.shape[0],2,3)*255).numpy().astype(np.uint8) | |
name='bike'+opt.replace(':','_')+'.png' | |
output=cv2.imread('bike'+opt.replace(':','_')+'.png') | |
output=cv2.resize(output,(output.shape[1]*256//output.shape[0],256)) | |
return output[:,:,::-1]#np.concatenate([img1,white,img2,white,img3,white,img4,white,img5],axis=1) | |
with gr.Blocks() as demo: | |
gr.Markdown("image cropping") | |
#with gr.Tab("Component test"): | |
with gr.Row(): | |
with gr.Column(): | |
#slider1 = gr.Slider(2, 20, value=2, label="Count", info="Choose betwen 2 and 20") | |
#drop1 = gr.Dropdown(["cat", "dog", "bird"], label="Animal", info="Will add more animals later!") | |
#checklist1 = gr.CheckboxGroup(["4:3", "3:4", "16:9"], label="Shape", info="The shape of cropped image") | |
radio2 = gr.Radio(["9:16", "3:4","1:1","4:3", "16:9","circle"], value="3:4",label="Shape", info="The shape of cropped image") | |
#radio1 = gr.Radio(["park", "zoo", "road"], label="Location", info="Where did they go?") | |
src_img1 = gr.Image() | |
exp=gr.Examples(["bike.png","img2.png"],src_img1) | |
bottom1 = gr.Button(label="cropping component") | |
out1 = gr.Image()#gr.Textbox() | |
bottom1.click(sepia, inputs=[radio2,src_img1], outputs=out1) | |
#gr.Image(shape=(224, 2)) | |
#demo = gr.Interface(sepia, inputs=["image","image","image","image","image"], outputs=["image","image","image","image","image"])#gr.Interface(sepia, gr.Image(shape=(200, 200)), "image") | |
#demo = gr.Interface(sepia, inputs=["image"], outputs=["image"]) | |
demo.launch(debug=True) |