Ariamehr commited on
Commit
4916b73
1 Parent(s): 6fab9d9

Upload 4 files

Browse files
Files changed (4) hide show
  1. app.py +224 -0
  2. banner.html +68 -0
  3. gitignore +2 -0
  4. tips.html +24 -0
app.py ADDED
@@ -0,0 +1,224 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import colorsys
2
+ import os
3
+
4
+ import gradio as gr
5
+ import matplotlib.colors as mcolors
6
+ import numpy as np
7
+ import torch
8
+ from gradio.themes.utils import sizes
9
+ from matplotlib import pyplot as plt
10
+ from matplotlib.patches import Patch
11
+ from PIL import Image
12
+ from torchvision import transforms
13
+
14
+ # ----------------- HELPER FUNCTIONS ----------------- #
15
+
16
+ ASSETS_DIR = os.path.join(os.path.dirname(__file__), "assets")
17
+
18
+ LABELS_TO_IDS = {
19
+ "Background": 0,
20
+ "Apparel": 1,
21
+ "Face Neck": 2,
22
+ "Hair": 3,
23
+ "Left Foot": 4,
24
+ "Left Hand": 5,
25
+ "Left Lower Arm": 6,
26
+ "Left Lower Leg": 7,
27
+ "Left Shoe": 8,
28
+ "Left Sock": 9,
29
+ "Left Upper Arm": 10,
30
+ "Left Upper Leg": 11,
31
+ "Lower Clothing": 12,
32
+ "Right Foot": 13,
33
+ "Right Hand": 14,
34
+ "Right Lower Arm": 15,
35
+ "Right Lower Leg": 16,
36
+ "Right Shoe": 17,
37
+ "Right Sock": 18,
38
+ "Right Upper Arm": 19,
39
+ "Right Upper Leg": 20,
40
+ "Torso": 21,
41
+ "Upper Clothing": 22,
42
+ "Lower Lip": 23,
43
+ "Upper Lip": 24,
44
+ "Lower Teeth": 25,
45
+ "Upper Teeth": 26,
46
+ "Tongue": 27,
47
+ }
48
+
49
+
50
+ def get_palette(num_cls):
51
+ palette = [0] * (256 * 3)
52
+ palette[0:3] = [0, 0, 0]
53
+
54
+ for j in range(1, num_cls):
55
+ hue = (j - 1) / (num_cls - 1)
56
+ saturation = 1.0
57
+ value = 1.0 if j % 2 == 0 else 0.5
58
+ rgb = colorsys.hsv_to_rgb(hue, saturation, value)
59
+ r, g, b = [int(x * 255) for x in rgb]
60
+ palette[j * 3 : j * 3 + 3] = [r, g, b]
61
+
62
+ return palette
63
+
64
+
65
+ def create_colormap(palette):
66
+ colormap = np.array(palette).reshape(-1, 3) / 255.0
67
+ return mcolors.ListedColormap(colormap)
68
+
69
+
70
+ def visualize_mask_with_overlay(img: Image.Image, mask: Image.Image, labels_to_ids: dict[str, int], alpha=0.5):
71
+ img_np = np.array(img.convert("RGB"))
72
+ mask_np = np.array(mask)
73
+
74
+ num_cls = len(labels_to_ids)
75
+ palette = get_palette(num_cls)
76
+ colormap = create_colormap(palette)
77
+
78
+ overlay = np.zeros((*mask_np.shape, 3), dtype=np.uint8)
79
+ for label, idx in labels_to_ids.items():
80
+ if idx != 0:
81
+ overlay[mask_np == idx] = np.array(colormap(idx)[:3]) * 255
82
+
83
+ blended = Image.fromarray(np.uint8(img_np * (1 - alpha) + overlay * alpha))
84
+
85
+ return blended
86
+
87
+
88
+ def create_legend_image(labels_to_ids: dict[str, int], filename="legend.png"):
89
+ num_cls = len(labels_to_ids)
90
+ palette = get_palette(num_cls)
91
+ colormap = create_colormap(palette)
92
+
93
+ fig, ax = plt.subplots(figsize=(4, 6), facecolor="white")
94
+
95
+ ax.axis("off")
96
+
97
+ legend_elements = [
98
+ Patch(facecolor=colormap(i), edgecolor="black", label=label)
99
+ for label, i in sorted(labels_to_ids.items(), key=lambda x: x[1])
100
+ ]
101
+
102
+ plt.title("Legend", fontsize=16, fontweight="bold", pad=20)
103
+
104
+ legend = ax.legend(
105
+ handles=legend_elements,
106
+ loc="center",
107
+ bbox_to_anchor=(0.5, 0.5),
108
+ ncol=2,
109
+ frameon=True,
110
+ fancybox=True,
111
+ shadow=True,
112
+ fontsize=10,
113
+ title_fontsize=12,
114
+ borderpad=1,
115
+ labelspacing=1.2,
116
+ handletextpad=0.5,
117
+ handlelength=1.5,
118
+ columnspacing=1.5,
119
+ )
120
+
121
+ legend.get_frame().set_facecolor("#FAFAFA")
122
+ legend.get_frame().set_edgecolor("gray")
123
+
124
+ # Adjust layout and save
125
+ plt.tight_layout()
126
+ plt.savefig(filename, dpi=300, bbox_inches="tight")
127
+ plt.close()
128
+
129
+
130
+ # create_legend_image(LABELS_TO_IDS, filename=os.path.join(ASSETS_DIR, "legend.png"))
131
+
132
+
133
+ # ----------------- MODEL ----------------- #
134
+
135
+ URL = "https://huggingface.co/facebook/sapiens/resolve/main/sapiens_lite_host/torchscript/seg/checkpoints/sapiens_0.3b/sapiens_0.3b_goliath_best_goliath_mIoU_7673_epoch_194_torchscript.pt2?download=true"
136
+ CHECKPOINTS_DIR = os.path.join(ASSETS_DIR, "checkpoints")
137
+ model_path = os.path.join(CHECKPOINTS_DIR, "sapiens_0.3b_goliath_best_goliath_mIoU_7673_epoch_194_torchscript.pt2")
138
+
139
+ if not os.path.exists(model_path):
140
+ os.makedirs(CHECKPOINTS_DIR, exist_ok=True)
141
+ import requests
142
+
143
+ response = requests.get(URL)
144
+ with open(model_path, "wb") as file:
145
+ file.write(response.content)
146
+
147
+ model = torch.jit.load(model_path)
148
+ model.eval()
149
+
150
+
151
+ @torch.no_grad()
152
+ def run_model(input_tensor, height, width):
153
+ output = model(input_tensor)
154
+ output = torch.nn.functional.interpolate(output, size=(height, width), mode="bilinear", align_corners=False)
155
+ _, preds = torch.max(output, 1)
156
+ return preds
157
+
158
+
159
+ transform_fn = transforms.Compose(
160
+ [
161
+ transforms.Resize((1024, 768)),
162
+ transforms.ToTensor(),
163
+ transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
164
+ ]
165
+ )
166
+ # ----------------- CORE FUNCTION ----------------- #
167
+
168
+
169
+ def segment(image: Image.Image) -> Image.Image:
170
+ input_tensor = transform_fn(image).unsqueeze(0)
171
+ preds = run_model(input_tensor, height=image.height, width=image.width)
172
+ mask = preds.squeeze(0).cpu().numpy()
173
+ mask_image = Image.fromarray(mask.astype("uint8"))
174
+ blended_image = visualize_mask_with_overlay(image, mask_image, LABELS_TO_IDS, alpha=0.5)
175
+ return blended_image
176
+
177
+
178
+ # ----------------- GRADIO UI ----------------- #
179
+
180
+
181
+ with open("banner.html", "r") as file:
182
+ banner = file.read()
183
+ with open("tips.html", "r") as file:
184
+ tips = file.read()
185
+
186
+ CUSTOM_CSS = """
187
+ .image-container img {
188
+ max-width: 512px;
189
+ max-height: 512px;
190
+ margin: 0 auto;
191
+ border-radius: 0px;
192
+ .gradio-container {background-color: #fafafa}
193
+ """
194
+
195
+ with gr.Blocks(css=CUSTOM_CSS, theme=gr.themes.Monochrome(radius_size=sizes.radius_md)) as demo:
196
+ gr.HTML(banner)
197
+ gr.HTML(tips)
198
+ with gr.Row():
199
+ with gr.Column():
200
+ input_image = gr.Image(label="Input Image", type="pil", format="png")
201
+
202
+ example_model = gr.Examples(
203
+ inputs=input_image,
204
+ examples_per_page=10,
205
+ examples=[
206
+ os.path.join(ASSETS_DIR, "examples", img)
207
+ for img in os.listdir(os.path.join(ASSETS_DIR, "examples"))
208
+ ],
209
+ )
210
+ with gr.Column():
211
+ result_image = gr.Image(label="Segmentation Result", format="png")
212
+ run_button = gr.Button("Run")
213
+
214
+ gr.Image(os.path.join(ASSETS_DIR, "legend.png"), label="Legend", type="filepath")
215
+
216
+ run_button.click(
217
+ fn=segment,
218
+ inputs=[input_image],
219
+ outputs=[result_image],
220
+ )
221
+
222
+
223
+ if __name__ == "__main__":
224
+ demo.launch(share=False)
banner.html ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <div style="
2
+ display: flex;
3
+ flex-direction: column;
4
+ justify-content: center;
5
+ align-items: center;
6
+ text-align: center;
7
+ background: linear-gradient(45deg, #1a1a1a 0%, #333333 100%);
8
+ padding: 24px;
9
+ gap: 24px;
10
+ border-radius: 8px;
11
+ ">
12
+ <div style="display: flex; gap: 8px;">
13
+ <h1 style="
14
+ font-size: 48px;
15
+ color: #fafafa;
16
+ margin: 0;
17
+ font-family: 'Trebuchet MS', 'Lucida Sans Unicode', 'Lucida Grande',
18
+ 'Lucida Sans', Arial, sans-serif;
19
+ ">
20
+ Sapiens 0.3B: Body-part Segmentation
21
+ </h1>
22
+
23
+
24
+ </div>
25
+
26
+ <p style="
27
+ margin: 0;
28
+ line-height: 1.6rem;
29
+ font-size: 16px;
30
+ color: #fafafa;
31
+ opacity: 0.8;
32
+ ">
33
+ <a href="https://about.meta.com/realitylabs/codecavatars/sapiens/" target="_blank">Sapiens</a> is a human-centric
34
+ family of foundational models trained by Meta Reality Labs. <br />
35
+ This Space is brought to you by FASHN AI, for your convenience, to showcase the capabilities of Sapiens for
36
+ body-part Segmentation.
37
+
38
+ </p>
39
+
40
+ <div style="
41
+ display: flex;
42
+ justify-content: center;
43
+ align-items: center;
44
+ text-align: center;
45
+ ">
46
+ <a href="https://fashn.ai"><img
47
+ src="https://custom-icon-badges.demolab.com/badge/FASHN_AI-333333?style=for-the-badge&logo=fashn"
48
+ alt="FASHN AI" /></a>
49
+ <a href="https://github.com/fashn-AI"><img
50
+ src="https://img.shields.io/badge/github-%23121011.svg?style=for-the-badge&logo=github&logoColor=white"
51
+ alt="Github" /></a>
52
+ <a href="https://www.linkedin.com/company/fashn">
53
+ <img src="https://img.shields.io/badge/linkedin-%230077B5.svg?style=for-the-badge&logo=linkedin&logoColor=white"
54
+ alt="LinkedIn" />
55
+ </a>
56
+
57
+ <a href="https://x.com/fashn_ai"><img
58
+ src="https://img.shields.io/badge/@fashn_ai-%23000000.svg?style=for-the-badge&logo=X&logoColor=white"
59
+ alt="X" /></a>
60
+ <a href="https://www.instagram.com/fashn.ai/"><img
61
+ src="https://img.shields.io/badge/Fashn.ai-%23E4405F.svg?style=for-the-badge&logo=Instagram&logoColor=white"
62
+ alt="Instagram" /></a>
63
+ <a href="https://discord.gg/zfqzkGBxE5">
64
+ <img src="https://img.shields.io/badge/fashn_ai-%235865F2.svg?style=for-the-badge&logo=discord&logoColor=white"
65
+ alt="Discord" />
66
+ </a>
67
+ </div>
68
+ </div>
gitignore ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ .DS_Store
2
+ *.pt2
tips.html ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <div style="
2
+ padding: 12px;
3
+ border: 1px solid #333333;
4
+ border-radius: 8px;
5
+ text-align: center;
6
+ display: flex;
7
+ flex-direction: column;
8
+ gap: 8px;
9
+ ">
10
+ <b style="font-size: 18px;"> ❣️ Tips for successful segmentations</b>
11
+
12
+ <ul style="
13
+ display: flex;
14
+ gap: 12px;
15
+ justify-content: center;
16
+ li {
17
+ margin: 0;
18
+ }
19
+ ">
20
+ <li>3:4 aspect ratio</li>
21
+ <li>768x1024 (width x height) resolution</li>
22
+
23
+ </ul>
24
+ </div>