Iiniciado el proyecto de mariposas
Browse files- .gitignore +1 -0
- app.py +63 -0
- assets/app.py +70 -0
- packages.txt +1 -0
- requirements.txt +5 -0
- utils.py +22 -0
.gitignore
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
venv/
|
app.py
ADDED
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
|
3 |
+
from utils import carga_modelo, genera
|
4 |
+
|
5 |
+
## Página principal
|
6 |
+
st.title("Butterfly GAN (GAN de mariposas)")
|
7 |
+
st.write(
|
8 |
+
"Modelo Light-GAN entrenado con 1000 imágenes de mariposas tomadas de la colección del Museo Smithsonian."
|
9 |
+
)
|
10 |
+
|
11 |
+
## Barra lateral
|
12 |
+
st.sidebar.subheader("¡Esta mariposa no existe! Ni en América Latina 🤯.")
|
13 |
+
st.sidebar.image("assets/logo.png", width=200)
|
14 |
+
st.sidebar.caption(
|
15 |
+
f"[Modelo](https://huggingface.co/ceyda/butterfly_cropped_uniq1K_512) y [Dataset](https://huggingface.co/datasets/huggan/smithsonian_butterflies_subset) usados."
|
16 |
+
)
|
17 |
+
st.sidebar.caption(f"*Disclaimers:*")
|
18 |
+
st.sidebar.caption(
|
19 |
+
"* Este demo es una versión simplificada del creado por [Ceyda Cinarel](https://github.com/cceyda) y [Jonathan Whitaker](https://datasciencecastnet.home.blog/) ([link](https://huggingface.co/spaces/huggan/butterfly-gan)) durante el hackathon [HugGan](https://github.com/huggingface/community-events). Cualquier error se atribuye a [Omar Espejel](https://twitter.com/espejelomar)."
|
20 |
+
)
|
21 |
+
st.sidebar.caption(
|
22 |
+
"* Modelo basado en el [paper](https://openreview.net/forum?id=1Fqg133qRaI) *Towards Faster and Stabilized GAN Training for High-fidelity Few-shot Image Synthesis*."
|
23 |
+
)
|
24 |
+
|
25 |
+
## Cargamos modelo
|
26 |
+
repo_id = "ceyda/butterfly_cropped_uniq1K_512"
|
27 |
+
version_modelo = "57d36a15546909557d9f967f47713236c8288838"
|
28 |
+
modelo_gan = carga_modelo(repo_id, version_modelo)
|
29 |
+
|
30 |
+
## Generamos 4 mariposas
|
31 |
+
n_mariposas = 4
|
32 |
+
|
33 |
+
## Función que genera mariposas y lo guarda como un estado de la sesión
|
34 |
+
def corre():
|
35 |
+
#bolita e espera que gira indcándote que se está generando
|
36 |
+
with st.spinner("Generando, espera un poco..."):
|
37 |
+
#mientras generamos nuestras mariposas con el modelo gan
|
38 |
+
ims = genera(modelo_gan, n_mariposas)
|
39 |
+
#guardamos las imágenes en un state session llamado ims
|
40 |
+
st.session_state["ims"] = ims
|
41 |
+
|
42 |
+
|
43 |
+
## Si no hay una imagen generada entonces generala
|
44 |
+
if "ims" not in st.session_state:
|
45 |
+
st.session_state["ims"] = None
|
46 |
+
corre()
|
47 |
+
|
48 |
+
## ims contiene las imágenes generadas
|
49 |
+
ims = st.session_state["ims"]
|
50 |
+
|
51 |
+
## Si la usuaria da click en el botón entonces corremos la función genera()
|
52 |
+
#Creamos un botón
|
53 |
+
corre_boton = st.button(
|
54 |
+
"Genera mariposas, porfa.",
|
55 |
+
on_click=corre,
|
56 |
+
help="Estamos en pleno vuelo, puede tardar.",
|
57 |
+
)
|
58 |
+
#Si no está vacía creamos columnas con n mariposas
|
59 |
+
if ims is not None:
|
60 |
+
cols = st.columns(n_mariposas)
|
61 |
+
for j, im in enumerate(ims):
|
62 |
+
i = j % n_mariposas
|
63 |
+
cols[i].image(im, use_column_width=True)
|
assets/app.py
ADDED
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from PIL import Image
|
3 |
+
import numpy as np
|
4 |
+
import cv2
|
5 |
+
from huggingface_hub import from_pretrained_keras
|
6 |
+
|
7 |
+
st.header("Segmentación de dientes con rayos X")
|
8 |
+
|
9 |
+
st.markdown('''
|
10 |
+
|
11 |
+
Hola estudiantes de Platzi 🚀. Este modelo usan UNet para segmentar imágenes
|
12 |
+
de dientos en rayos X. Se utila un modelo de Keras importado con la función
|
13 |
+
`huggingface_hub.from_pretrained_keras`. Recuerda que el Hub de Hugging Face está integrado
|
14 |
+
con muchas librerías como Keras, scikit-learn, fastai y otras.
|
15 |
+
|
16 |
+
El modelo fue creado por [SerdarHelli](https://huggingface.co/SerdarHelli/Segmentation-of-Teeth-in-Panoramic-X-ray-Image-Using-U-Net).
|
17 |
+
|
18 |
+
''')
|
19 |
+
|
20 |
+
model_id = "SerdarHelli/Segmentation-of-Teeth-in-Panoramic-X-ray-Image-Using-U-Net"
|
21 |
+
model=from_pretrained_keras(model_id)
|
22 |
+
|
23 |
+
## Si una imagen tiene más de un canal entonces se convierte a escala de grises (1 canal)
|
24 |
+
def convertir_one_channel(img):
|
25 |
+
if len(img.shape)>2:
|
26 |
+
img= cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
27 |
+
return img
|
28 |
+
else:
|
29 |
+
return img
|
30 |
+
|
31 |
+
def convertir_rgb(img):
|
32 |
+
if len(img.shape)==2:
|
33 |
+
img= cv2.cvtColor(img,cv2.COLOR_GRAY2RGB)
|
34 |
+
return img
|
35 |
+
else:
|
36 |
+
return img
|
37 |
+
|
38 |
+
|
39 |
+
image_file = st.file_uploader("Sube aquí tu imagen.", type=["png","jpg","jpeg"])
|
40 |
+
|
41 |
+
|
42 |
+
if image_file is not None:
|
43 |
+
|
44 |
+
img= Image.open(image_file)
|
45 |
+
|
46 |
+
st.image(img,width=850)
|
47 |
+
|
48 |
+
img=np.asarray(img)
|
49 |
+
|
50 |
+
img_cv=convertir_one_channel(img)
|
51 |
+
img_cv=cv2.resize(img_cv,(512,512), interpolation=cv2.INTER_LANCZOS4)
|
52 |
+
img_cv=np.float32(img_cv/255)
|
53 |
+
|
54 |
+
img_cv=np.reshape(img_cv,(1,512,512,1))
|
55 |
+
prediction=model.predict(img_cv)
|
56 |
+
predicted=prediction[0]
|
57 |
+
predicted = cv2.resize(predicted, (img.shape[1],img.shape[0]), interpolation=cv2.INTER_LANCZOS4)
|
58 |
+
mask=np.uint8(predicted*255)#
|
59 |
+
_, mask = cv2.threshold(mask, thresh=0, maxval=255, type=cv2.THRESH_BINARY+cv2.THRESH_OTSU)
|
60 |
+
kernel =( np.ones((5,5), dtype=np.float32))
|
61 |
+
mask=cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel,iterations=1 )
|
62 |
+
mask=cv2.morphologyEx(mask, cv2.MORPH_CLOSE, kernel,iterations=1 )
|
63 |
+
cnts,hieararch=cv2.findContours(mask,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
|
64 |
+
output = cv2.drawContours(convertir_one_channel(img), cnts, -1, (255, 0, 0) , 3)
|
65 |
+
|
66 |
+
|
67 |
+
if output is not None :
|
68 |
+
st.subheader("Segmentación:")
|
69 |
+
st.write(output.shape)
|
70 |
+
st.image(output,width=850)
|
packages.txt
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
libgl1
|
requirements.txt
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
git+https://github.com/huggingface/community-events.git@3fea10c5d5a50c69f509e34cd580fe9139905d04#egg=huggan
|
2 |
+
transformers
|
3 |
+
faiss-cpu
|
4 |
+
paddlehub
|
5 |
+
paddlepaddle
|
utils.py
ADDED
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Funciones que ayudarñán en el resto de la app
|
2 |
+
|
3 |
+
import numpy as mp
|
4 |
+
import torch
|
5 |
+
from huggan.pytorch.lightweigth_gan.lightweigth_gan import lightweightGAN #par generar imágenes
|
6 |
+
|
7 |
+
## Cargamos el modelo desde el Hub de Hugging Face
|
8 |
+
def carga_modelo(model_name="ceyda/butterfly_cropped_uniq1K_512", model_version=None):
|
9 |
+
#creamos nuestra gan y descargamos el modelo
|
10 |
+
gan = LightweightGAN.from_pretrained(model_name, version=model_version)
|
11 |
+
#ponemos nuestra gan en forma de evaluacion
|
12 |
+
gan.eval()
|
13 |
+
return gan
|
14 |
+
|
15 |
+
|
16 |
+
## Usamos el modelo GAN para generar imágenes, y le decimos que el tamaño será 1
|
17 |
+
def genera(gan, batch_size=1):
|
18 |
+
#utilizando pytorch vamos a llamar a nuestra gan
|
19 |
+
with torch.no_grad():
|
20 |
+
ims = gan.G(torch.randn(batch_size, gan.latent_dim)).clamp_(0.0, 1.0) * 255
|
21 |
+
ims = ims.permute(0, 2, 3, 1).detach().cpu().numpy().astype(np.uint8)
|
22 |
+
return ims
|