Spaces:
Runtime error
Runtime error
reparando errores
Browse files
app.py
CHANGED
@@ -1,34 +1,59 @@
|
|
1 |
import streamlit as st
|
2 |
|
3 |
-
from utils import
|
4 |
-
|
5 |
-
|
6 |
-
st.
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
if "ims" not in st.session_state:
|
20 |
-
st.session_state["ims"]=None
|
21 |
-
|
22 |
|
23 |
-
ims
|
|
|
24 |
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
|
|
|
|
29 |
|
30 |
if ims is not None:
|
31 |
-
cols=st.columns(
|
32 |
for j, im in enumerate(ims):
|
33 |
-
i=j%
|
34 |
-
cols[i].image(im,use_column_width=True)
|
|
|
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 |
+
with st.spinner("Generando, espera un poco..."):
|
36 |
+
ims = genera(modelo_gan, n_mariposas)
|
37 |
+
st.session_state["ims"] = ims
|
38 |
+
|
39 |
+
|
40 |
+
## Si no hay una imagen generada entonces generala
|
41 |
if "ims" not in st.session_state:
|
42 |
+
st.session_state["ims"] = None
|
43 |
+
corre()
|
44 |
|
45 |
+
## ims contiene las imágenes generadas
|
46 |
+
ims = st.session_state["ims"]
|
47 |
|
48 |
+
## Si la usuaria da click en el botón entonces corremos la función genera()
|
49 |
+
corre_boton = st.button(
|
50 |
+
"Genera mariposas, porfa.",
|
51 |
+
on_click=corre,
|
52 |
+
help="Estamos en pleno vuelo, puede tardar.",
|
53 |
+
)
|
54 |
|
55 |
if ims is not None:
|
56 |
+
cols = st.columns(n_mariposas)
|
57 |
for j, im in enumerate(ims):
|
58 |
+
i = j % n_mariposas
|
59 |
+
cols[i].image(im, use_column_width=True)
|
utils.py
CHANGED
@@ -2,13 +2,17 @@ import numpy as np
|
|
2 |
import torch
|
3 |
from huggan.pytorch.lightweight_gan.lightweight_gan import LightweightGAN
|
4 |
|
5 |
-
|
6 |
-
|
|
|
|
|
7 |
gan.eval()
|
8 |
return gan
|
9 |
|
10 |
-
|
|
|
|
|
11 |
with torch.no_grad():
|
12 |
-
ims=gan.G(torch.randn(batch_size,gan.latent_dim)).
|
13 |
-
ims=ims.permute(0,2,3,1).
|
14 |
-
|
|
|
2 |
import torch
|
3 |
from huggan.pytorch.lightweight_gan.lightweight_gan import LightweightGAN
|
4 |
|
5 |
+
|
6 |
+
## Cargamos el modelo desde el Hub de Hugging Face
|
7 |
+
def carga_modelo(model_name="ceyda/butterfly_cropped_uniq1K_512", model_version=None):
|
8 |
+
gan = LightweightGAN.from_pretrained(model_name, version=model_version)
|
9 |
gan.eval()
|
10 |
return gan
|
11 |
|
12 |
+
|
13 |
+
## Usamos el modelo GAN para generar imágenes
|
14 |
+
def genera(gan, batch_size=1):
|
15 |
with torch.no_grad():
|
16 |
+
ims = gan.G(torch.randn(batch_size, gan.latent_dim)).clamp_(0.0, 1.0) * 255
|
17 |
+
ims = ims.permute(0, 2, 3, 1).detach().cpu().numpy().astype(np.uint8)
|
18 |
+
return ims
|