Spaces:
Sleeping
Sleeping
Commit
·
b1ae84d
1
Parent(s):
1b58573
Update
Browse files- README.md +2 -2
- Utils/dbimutils.py +54 -0
- app.py +134 -115
- miku.jpg +0 -0
- miku2.jpg +0 -0
- power.jpg +0 -0
- requirements.txt +3 -3
README.md
CHANGED
|
@@ -1,10 +1,10 @@
|
|
| 1 |
---
|
| 2 |
-
title:
|
| 3 |
emoji: 💬
|
| 4 |
colorFrom: blue
|
| 5 |
colorTo: red
|
| 6 |
sdk: gradio
|
| 7 |
-
sdk_version: 3.
|
| 8 |
app_file: app.py
|
| 9 |
pinned: false
|
| 10 |
duplicated_from: NoCrypt/DeepDanbooru_string
|
|
|
|
| 1 |
---
|
| 2 |
+
title: WaifuDiffusion v1.4 Tags
|
| 3 |
emoji: 💬
|
| 4 |
colorFrom: blue
|
| 5 |
colorTo: red
|
| 6 |
sdk: gradio
|
| 7 |
+
sdk_version: 3.6
|
| 8 |
app_file: app.py
|
| 9 |
pinned: false
|
| 10 |
duplicated_from: NoCrypt/DeepDanbooru_string
|
Utils/dbimutils.py
ADDED
|
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DanBooru IMage Utility functions
|
| 2 |
+
|
| 3 |
+
import cv2
|
| 4 |
+
import numpy as np
|
| 5 |
+
from PIL import Image
|
| 6 |
+
|
| 7 |
+
|
| 8 |
+
def smart_imread(img, flag=cv2.IMREAD_UNCHANGED):
|
| 9 |
+
if img.endswith(".gif"):
|
| 10 |
+
img = Image.open(img)
|
| 11 |
+
img = img.convert("RGB")
|
| 12 |
+
img = cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR)
|
| 13 |
+
else:
|
| 14 |
+
img = cv2.imread(img, flag)
|
| 15 |
+
return img
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
def smart_24bit(img):
|
| 19 |
+
if img.dtype is np.dtype(np.uint16):
|
| 20 |
+
img = (img / 257).astype(np.uint8)
|
| 21 |
+
|
| 22 |
+
if len(img.shape) == 2:
|
| 23 |
+
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
|
| 24 |
+
elif img.shape[2] == 4:
|
| 25 |
+
trans_mask = img[:, :, 3] == 0
|
| 26 |
+
img[trans_mask] = [255, 255, 255, 255]
|
| 27 |
+
img = cv2.cvtColor(img, cv2.COLOR_BGRA2BGR)
|
| 28 |
+
return img
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
def make_square(img, target_size):
|
| 32 |
+
old_size = img.shape[:2]
|
| 33 |
+
desired_size = max(old_size)
|
| 34 |
+
desired_size = max(desired_size, target_size)
|
| 35 |
+
|
| 36 |
+
delta_w = desired_size - old_size[1]
|
| 37 |
+
delta_h = desired_size - old_size[0]
|
| 38 |
+
top, bottom = delta_h // 2, delta_h - (delta_h // 2)
|
| 39 |
+
left, right = delta_w // 2, delta_w - (delta_w // 2)
|
| 40 |
+
|
| 41 |
+
color = [255, 255, 255]
|
| 42 |
+
new_im = cv2.copyMakeBorder(
|
| 43 |
+
img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color
|
| 44 |
+
)
|
| 45 |
+
return new_im
|
| 46 |
+
|
| 47 |
+
|
| 48 |
+
def smart_resize(img, size):
|
| 49 |
+
# Assumes the image has already gone through make_square
|
| 50 |
+
if img.shape[0] > size:
|
| 51 |
+
img = cv2.resize(img, (size, size), interpolation=cv2.INTER_AREA)
|
| 52 |
+
elif img.shape[0] < size:
|
| 53 |
+
img = cv2.resize(img, (size, size), interpolation=cv2.INTER_CUBIC)
|
| 54 |
+
return img
|
app.py
CHANGED
|
@@ -4,135 +4,164 @@ from __future__ import annotations
|
|
| 4 |
|
| 5 |
import argparse
|
| 6 |
import functools
|
| 7 |
-
import os
|
| 8 |
import html
|
| 9 |
-
import
|
| 10 |
-
import tarfile
|
| 11 |
|
| 12 |
-
import deepdanbooru as dd
|
| 13 |
import gradio as gr
|
| 14 |
import huggingface_hub
|
| 15 |
import numpy as np
|
| 16 |
-
import
|
| 17 |
-
import
|
| 18 |
import piexif
|
| 19 |
import piexif.helper
|
|
|
|
|
|
|
|
|
|
| 20 |
|
| 21 |
-
TITLE =
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
|
| 23 |
-
|
| 24 |
-
MODEL_REPO =
|
| 25 |
-
MODEL_FILENAME =
|
| 26 |
-
LABEL_FILENAME =
|
| 27 |
|
| 28 |
|
| 29 |
def parse_args() -> argparse.Namespace:
|
| 30 |
parser = argparse.ArgumentParser()
|
| 31 |
-
parser.add_argument(
|
| 32 |
-
parser.add_argument(
|
| 33 |
-
parser.add_argument(
|
| 34 |
-
parser.add_argument('--live', action='store_true')
|
| 35 |
-
parser.add_argument('--share', action='store_true')
|
| 36 |
-
parser.add_argument('--port', type=int)
|
| 37 |
-
parser.add_argument('--disable-queue',
|
| 38 |
-
dest='enable_queue',
|
| 39 |
-
action='store_false')
|
| 40 |
-
parser.add_argument('--allow-flagging', type=str, default='never')
|
| 41 |
return parser.parse_args()
|
| 42 |
|
| 43 |
|
| 44 |
-
def
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
'images.tar.gz',
|
| 50 |
-
repo_type='dataset',
|
| 51 |
-
use_auth_token=TOKEN)
|
| 52 |
-
with tarfile.open(path) as f:
|
| 53 |
-
f.extractall()
|
| 54 |
-
return sorted(image_dir.glob('*'))
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
def load_model() -> tf.keras.Model:
|
| 58 |
-
path = huggingface_hub.hf_hub_download(MODEL_REPO,
|
| 59 |
-
MODEL_FILENAME,
|
| 60 |
-
use_auth_token=TOKEN)
|
| 61 |
-
model = tf.keras.models.load_model(path)
|
| 62 |
return model
|
| 63 |
|
| 64 |
|
| 65 |
def load_labels() -> list[str]:
|
| 66 |
-
path = huggingface_hub.hf_hub_download(
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
|
| 73 |
def plaintext_to_html(text):
|
| 74 |
-
text =
|
|
|
|
|
|
|
| 75 |
return text
|
| 76 |
|
| 77 |
-
|
| 78 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 79 |
rawimage = image
|
| 80 |
-
_, height, width, _ = model.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 81 |
image = np.asarray(image)
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
image =
|
| 87 |
-
image =
|
| 88 |
-
image = image
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 100 |
items = rawimage.info
|
| 101 |
-
geninfo =
|
| 102 |
-
|
| 103 |
if "exif" in rawimage.info:
|
| 104 |
exif = piexif.load(rawimage.info["exif"])
|
| 105 |
-
exif_comment = (exif or {}).get("Exif", {}).get(piexif.ExifIFD.UserComment, b
|
| 106 |
try:
|
| 107 |
exif_comment = piexif.helper.UserComment.load(exif_comment)
|
| 108 |
except ValueError:
|
| 109 |
-
exif_comment = exif_comment.decode(
|
| 110 |
-
|
| 111 |
-
items[
|
| 112 |
geninfo = exif_comment
|
| 113 |
-
|
| 114 |
-
for field in [
|
| 115 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 116 |
items.pop(field, None)
|
| 117 |
-
|
| 118 |
-
geninfo = items.get(
|
| 119 |
-
|
| 120 |
info = f"""
|
| 121 |
<p><h4>PNG Info</h4></p>
|
| 122 |
"""
|
| 123 |
for key, text in items.items():
|
| 124 |
-
info +=
|
|
|
|
| 125 |
<div>
|
| 126 |
<p><b>{plaintext_to_html(str(key))}</b></p>
|
| 127 |
<p>{plaintext_to_html(str(text))}</p>
|
| 128 |
</div>
|
| 129 |
-
""".strip()
|
| 130 |
-
|
|
|
|
|
|
|
| 131 |
if len(info) == 0:
|
| 132 |
message = "Nothing found in the image."
|
| 133 |
info = f"<div><p>{message}<p></div>"
|
| 134 |
-
|
| 135 |
-
return (a,c,res,info)
|
| 136 |
|
| 137 |
|
| 138 |
def main():
|
|
@@ -141,45 +170,35 @@ def main():
|
|
| 141 |
labels = load_labels()
|
| 142 |
|
| 143 |
func = functools.partial(predict, model=model, labels=labels)
|
| 144 |
-
func = functools.update_wrapper(func, predict)
|
| 145 |
|
| 146 |
gr.Interface(
|
| 147 |
-
func,
|
| 148 |
-
[
|
| 149 |
-
gr.
|
| 150 |
-
gr.
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
gr.outputs.Textbox(label='Output (string)'),
|
| 158 |
-
gr.outputs.Textbox(label='Output (raw string)'),
|
| 159 |
-
gr.outputs.Label(label='Output (label)'),
|
| 160 |
-
gr.outputs.HTML()
|
| 161 |
],
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
|
|
|
|
|
|
|
|
|
| 165 |
],
|
|
|
|
| 166 |
title=TITLE,
|
| 167 |
-
description=
|
| 168 |
-
|
| 169 |
-
|
| 170 |
-
Modified from [hysts/DeepDanbooru](https://huggingface.co/spaces/hysts/DeepDanbooru)
|
| 171 |
-
|
| 172 |
-
PNG Info code forked from [AUTOMATIC1111/stable-diffusion-webui](https://github.com/AUTOMATIC1111/stable-diffusion-webui)
|
| 173 |
-
''',
|
| 174 |
-
theme=args.theme,
|
| 175 |
-
allow_flagging=args.allow_flagging,
|
| 176 |
-
live=args.live,
|
| 177 |
).launch(
|
| 178 |
-
enable_queue=
|
| 179 |
-
server_port=args.port,
|
| 180 |
share=args.share,
|
| 181 |
)
|
| 182 |
|
| 183 |
|
| 184 |
-
if __name__ ==
|
| 185 |
main()
|
|
|
|
| 4 |
|
| 5 |
import argparse
|
| 6 |
import functools
|
|
|
|
| 7 |
import html
|
| 8 |
+
import os
|
|
|
|
| 9 |
|
|
|
|
| 10 |
import gradio as gr
|
| 11 |
import huggingface_hub
|
| 12 |
import numpy as np
|
| 13 |
+
import onnxruntime as rt
|
| 14 |
+
import pandas as pd
|
| 15 |
import piexif
|
| 16 |
import piexif.helper
|
| 17 |
+
import PIL.Image
|
| 18 |
+
|
| 19 |
+
from Utils import dbimutils
|
| 20 |
|
| 21 |
+
TITLE = "WaifuDiffusion v1.4 Tags"
|
| 22 |
+
DESCRIPTION = """
|
| 23 |
+
Demo for [SmilingWolf/wd-v1-4-vit-tagger](https://huggingface.co/SmilingWolf/wd-v1-4-vit-tagger) with "ready to copy" prompt and a prompt analyzer.
|
| 24 |
+
|
| 25 |
+
Modified from [NoCrypt/DeepDanbooru_string](https://huggingface.co/spaces/NoCrypt/DeepDanbooru_string)
|
| 26 |
+
Modified from [hysts/DeepDanbooru](https://huggingface.co/spaces/hysts/DeepDanbooru)
|
| 27 |
+
|
| 28 |
+
PNG Info code forked from [AUTOMATIC1111/stable-diffusion-webui](https://github.com/AUTOMATIC1111/stable-diffusion-webui)
|
| 29 |
+
"""
|
| 30 |
|
| 31 |
+
HF_TOKEN = os.environ["HF_TOKEN"]
|
| 32 |
+
MODEL_REPO = "SmilingWolf/wd-v1-4-vit-tagger"
|
| 33 |
+
MODEL_FILENAME = "ViTB16_11_07_2022_18h19m14s.onnx"
|
| 34 |
+
LABEL_FILENAME = "selected_tags.csv"
|
| 35 |
|
| 36 |
|
| 37 |
def parse_args() -> argparse.Namespace:
|
| 38 |
parser = argparse.ArgumentParser()
|
| 39 |
+
parser.add_argument("--score-slider-step", type=float, default=0.05)
|
| 40 |
+
parser.add_argument("--score-threshold", type=float, default=0.35)
|
| 41 |
+
parser.add_argument("--share", action="store_true")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
return parser.parse_args()
|
| 43 |
|
| 44 |
|
| 45 |
+
def load_model() -> rt.InferenceSession:
|
| 46 |
+
path = huggingface_hub.hf_hub_download(
|
| 47 |
+
MODEL_REPO, MODEL_FILENAME, use_auth_token=HF_TOKEN
|
| 48 |
+
)
|
| 49 |
+
model = rt.InferenceSession(path)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
return model
|
| 51 |
|
| 52 |
|
| 53 |
def load_labels() -> list[str]:
|
| 54 |
+
path = huggingface_hub.hf_hub_download(
|
| 55 |
+
MODEL_REPO, LABEL_FILENAME, use_auth_token=HF_TOKEN
|
| 56 |
+
)
|
| 57 |
+
df = pd.read_csv(path)["name"].tolist()
|
| 58 |
+
return df
|
| 59 |
+
|
| 60 |
|
| 61 |
def plaintext_to_html(text):
|
| 62 |
+
text = (
|
| 63 |
+
"<p>" + "<br>\n".join([f"{html.escape(x)}" for x in text.split("\n")]) + "</p>"
|
| 64 |
+
)
|
| 65 |
return text
|
| 66 |
|
| 67 |
+
|
| 68 |
+
def predict(
|
| 69 |
+
image: PIL.Image.Image,
|
| 70 |
+
score_threshold: float,
|
| 71 |
+
model: rt.InferenceSession,
|
| 72 |
+
labels: list[str],
|
| 73 |
+
):
|
| 74 |
rawimage = image
|
| 75 |
+
_, height, width, _ = model.get_inputs()[0].shape
|
| 76 |
+
|
| 77 |
+
# Alpha to white
|
| 78 |
+
image = image.convert("RGBA")
|
| 79 |
+
new_image = PIL.Image.new("RGBA", image.size, "WHITE")
|
| 80 |
+
new_image.paste(image, mask=image)
|
| 81 |
+
image = new_image.convert("RGB")
|
| 82 |
image = np.asarray(image)
|
| 83 |
+
|
| 84 |
+
# PIL RGB to OpenCV BGR
|
| 85 |
+
image = image[:, :, ::-1]
|
| 86 |
+
|
| 87 |
+
image = dbimutils.make_square(image, height)
|
| 88 |
+
image = dbimutils.smart_resize(image, height)
|
| 89 |
+
image = image.astype(np.float32)
|
| 90 |
+
image = np.expand_dims(image, 0)
|
| 91 |
+
|
| 92 |
+
input_name = model.get_inputs()[0].name
|
| 93 |
+
label_name = model.get_outputs()[0].name
|
| 94 |
+
probs = model.run([label_name], {input_name: image})[0]
|
| 95 |
+
|
| 96 |
+
labels = list(zip(labels, probs[0].astype(float)))
|
| 97 |
+
|
| 98 |
+
# First 4 labels are actually ratings: pick one with argmax
|
| 99 |
+
ratings_names = labels[:4]
|
| 100 |
+
rating = dict(ratings_names)
|
| 101 |
+
|
| 102 |
+
# Everything else is tags: pick any where prediction confidence > threshold
|
| 103 |
+
tags_names = labels[4:]
|
| 104 |
+
res = [x for x in tags_names if x[1] > score_threshold]
|
| 105 |
+
res = dict(res)
|
| 106 |
+
|
| 107 |
+
b = dict(sorted(res.items(), key=lambda item: item[1], reverse=True))
|
| 108 |
+
a = (
|
| 109 |
+
", ".join(list(b.keys()))
|
| 110 |
+
.replace("_", " ")
|
| 111 |
+
.replace("(", "\(")
|
| 112 |
+
.replace(")", "\)")
|
| 113 |
+
)
|
| 114 |
+
c = ", ".join(list(b.keys()))
|
| 115 |
+
|
| 116 |
items = rawimage.info
|
| 117 |
+
geninfo = ""
|
| 118 |
+
|
| 119 |
if "exif" in rawimage.info:
|
| 120 |
exif = piexif.load(rawimage.info["exif"])
|
| 121 |
+
exif_comment = (exif or {}).get("Exif", {}).get(piexif.ExifIFD.UserComment, b"")
|
| 122 |
try:
|
| 123 |
exif_comment = piexif.helper.UserComment.load(exif_comment)
|
| 124 |
except ValueError:
|
| 125 |
+
exif_comment = exif_comment.decode("utf8", errors="ignore")
|
| 126 |
+
|
| 127 |
+
items["exif comment"] = exif_comment
|
| 128 |
geninfo = exif_comment
|
| 129 |
+
|
| 130 |
+
for field in [
|
| 131 |
+
"jfif",
|
| 132 |
+
"jfif_version",
|
| 133 |
+
"jfif_unit",
|
| 134 |
+
"jfif_density",
|
| 135 |
+
"dpi",
|
| 136 |
+
"exif",
|
| 137 |
+
"loop",
|
| 138 |
+
"background",
|
| 139 |
+
"timestamp",
|
| 140 |
+
"duration",
|
| 141 |
+
]:
|
| 142 |
items.pop(field, None)
|
| 143 |
+
|
| 144 |
+
geninfo = items.get("parameters", geninfo)
|
| 145 |
+
|
| 146 |
info = f"""
|
| 147 |
<p><h4>PNG Info</h4></p>
|
| 148 |
"""
|
| 149 |
for key, text in items.items():
|
| 150 |
+
info += (
|
| 151 |
+
f"""
|
| 152 |
<div>
|
| 153 |
<p><b>{plaintext_to_html(str(key))}</b></p>
|
| 154 |
<p>{plaintext_to_html(str(text))}</p>
|
| 155 |
</div>
|
| 156 |
+
""".strip()
|
| 157 |
+
+ "\n"
|
| 158 |
+
)
|
| 159 |
+
|
| 160 |
if len(info) == 0:
|
| 161 |
message = "Nothing found in the image."
|
| 162 |
info = f"<div><p>{message}<p></div>"
|
| 163 |
+
|
| 164 |
+
return (a, c, rating, res, info)
|
| 165 |
|
| 166 |
|
| 167 |
def main():
|
|
|
|
| 170 |
labels = load_labels()
|
| 171 |
|
| 172 |
func = functools.partial(predict, model=model, labels=labels)
|
|
|
|
| 173 |
|
| 174 |
gr.Interface(
|
| 175 |
+
fn=func,
|
| 176 |
+
inputs=[
|
| 177 |
+
gr.Image(type="pil", label="Input"),
|
| 178 |
+
gr.Slider(
|
| 179 |
+
0,
|
| 180 |
+
1,
|
| 181 |
+
step=args.score_slider_step,
|
| 182 |
+
value=args.score_threshold,
|
| 183 |
+
label="Score Threshold",
|
| 184 |
+
),
|
|
|
|
|
|
|
|
|
|
|
|
|
| 185 |
],
|
| 186 |
+
outputs=[
|
| 187 |
+
gr.Textbox(label="Output (string)"),
|
| 188 |
+
gr.Textbox(label="Output (raw string)"),
|
| 189 |
+
gr.Label(label="Rating"),
|
| 190 |
+
gr.Label(label="Output (label)"),
|
| 191 |
+
gr.HTML(),
|
| 192 |
],
|
| 193 |
+
examples=[["power.jpg", 0.5]],
|
| 194 |
title=TITLE,
|
| 195 |
+
description=DESCRIPTION,
|
| 196 |
+
allow_flagging="never",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 197 |
).launch(
|
| 198 |
+
enable_queue=True,
|
|
|
|
| 199 |
share=args.share,
|
| 200 |
)
|
| 201 |
|
| 202 |
|
| 203 |
+
if __name__ == "__main__":
|
| 204 |
main()
|
miku.jpg
DELETED
|
Binary file (125 kB)
|
|
|
miku2.jpg
DELETED
|
Binary file (220 kB)
|
|
|
power.jpg
ADDED
|
requirements.txt
CHANGED
|
@@ -1,4 +1,4 @@
|
|
| 1 |
pillow>=9.0.0
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
|
|
|
| 1 |
pillow>=9.0.0
|
| 2 |
+
piexif>=1.1.3
|
| 3 |
+
onnxruntime>=1.12.0
|
| 4 |
+
opencv-python
|