Upload playground.ipynb
Browse files- playground.ipynb +167 -0
playground.ipynb
ADDED
@@ -0,0 +1,167 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cells": [
|
3 |
+
{
|
4 |
+
"cell_type": "code",
|
5 |
+
"execution_count": 65,
|
6 |
+
"metadata": {},
|
7 |
+
"outputs": [],
|
8 |
+
"source": [
|
9 |
+
"from PIL import Image\n",
|
10 |
+
"\n",
|
11 |
+
"import torch\n",
|
12 |
+
"from transformers import (\n",
|
13 |
+
" AutoModelForImageClassification,\n",
|
14 |
+
" AutoImageProcessor,\n",
|
15 |
+
")\n",
|
16 |
+
"import numpy as np"
|
17 |
+
]
|
18 |
+
},
|
19 |
+
{
|
20 |
+
"cell_type": "code",
|
21 |
+
"execution_count": 2,
|
22 |
+
"metadata": {},
|
23 |
+
"outputs": [],
|
24 |
+
"source": [
|
25 |
+
"MODEL_NAME = \"p1atdev/siglip-tagger-test-2\""
|
26 |
+
]
|
27 |
+
},
|
28 |
+
{
|
29 |
+
"cell_type": "code",
|
30 |
+
"execution_count": 44,
|
31 |
+
"metadata": {},
|
32 |
+
"outputs": [],
|
33 |
+
"source": [
|
34 |
+
"model = AutoModelForImageClassification.from_pretrained(\n",
|
35 |
+
" MODEL_NAME, torch_dtype=torch.bfloat16, trust_remote_code=True\n",
|
36 |
+
")\n",
|
37 |
+
"model.eval()\n",
|
38 |
+
"processor = AutoImageProcessor.from_pretrained(MODEL_NAME)"
|
39 |
+
]
|
40 |
+
},
|
41 |
+
{
|
42 |
+
"cell_type": "code",
|
43 |
+
"execution_count": 45,
|
44 |
+
"metadata": {},
|
45 |
+
"outputs": [],
|
46 |
+
"source": [
|
47 |
+
"image = Image.open(\"sample.jpg\")\n",
|
48 |
+
"inputs = processor(image, return_tensors=\"pt\").to(model.device, model.dtype)"
|
49 |
+
]
|
50 |
+
},
|
51 |
+
{
|
52 |
+
"cell_type": "code",
|
53 |
+
"execution_count": 70,
|
54 |
+
"metadata": {},
|
55 |
+
"outputs": [],
|
56 |
+
"source": [
|
57 |
+
"logits = model(**inputs).logits.detach().cpu().float()[0]\n",
|
58 |
+
"logits = np.clip(logits, 0.0, 1.0)"
|
59 |
+
]
|
60 |
+
},
|
61 |
+
{
|
62 |
+
"cell_type": "code",
|
63 |
+
"execution_count": 80,
|
64 |
+
"metadata": {},
|
65 |
+
"outputs": [],
|
66 |
+
"source": [
|
67 |
+
"results = {\n",
|
68 |
+
" model.config.id2label[i]: logit for i, logit in enumerate(logits) if logit > 0\n",
|
69 |
+
"}\n",
|
70 |
+
"results = sorted(results.items(), key=lambda x: x[1], reverse=True)"
|
71 |
+
]
|
72 |
+
},
|
73 |
+
{
|
74 |
+
"cell_type": "code",
|
75 |
+
"execution_count": 81,
|
76 |
+
"metadata": {},
|
77 |
+
"outputs": [
|
78 |
+
{
|
79 |
+
"name": "stdout",
|
80 |
+
"output_type": "stream",
|
81 |
+
"text": [
|
82 |
+
"1girl: 100.00%\n",
|
83 |
+
"outdoors: 100.00%\n",
|
84 |
+
"sky: 100.00%\n",
|
85 |
+
"solo: 100.00%\n",
|
86 |
+
"school uniform: 96.88%\n",
|
87 |
+
"skirt: 92.97%\n",
|
88 |
+
"day: 89.06%\n",
|
89 |
+
"cloud: 85.94%\n",
|
90 |
+
"scenery: 79.69%\n",
|
91 |
+
"pleated skirt: 72.27%\n",
|
92 |
+
"black hair: 66.80%\n",
|
93 |
+
"standing: 65.62%\n",
|
94 |
+
"sailor collar: 59.38%\n",
|
95 |
+
"sitting: 57.81%\n",
|
96 |
+
"long sleeves: 53.52%\n",
|
97 |
+
"serafuku: 53.12%\n",
|
98 |
+
"holding: 52.34%\n",
|
99 |
+
"tree: 47.46%\n",
|
100 |
+
"dress: 46.48%\n",
|
101 |
+
"shoes: 43.55%\n",
|
102 |
+
"building: 42.77%\n",
|
103 |
+
"neckerchief: 40.82%\n",
|
104 |
+
"short hair: 38.09%\n",
|
105 |
+
"water: 38.09%\n",
|
106 |
+
"cloudy sky: 37.30%\n",
|
107 |
+
"looking at viewer: 32.23%\n",
|
108 |
+
"long hair: 32.03%\n",
|
109 |
+
"brown eyes: 31.45%\n",
|
110 |
+
"plant: 31.05%\n",
|
111 |
+
"bag: 29.30%\n",
|
112 |
+
"railing: 29.10%\n",
|
113 |
+
"sunlight: 28.12%\n",
|
114 |
+
"from side: 27.73%\n",
|
115 |
+
"window: 27.54%\n",
|
116 |
+
"brown hair: 26.37%\n",
|
117 |
+
"white shirt: 25.78%\n",
|
118 |
+
"shirt: 25.39%\n",
|
119 |
+
"blue sky: 23.93%\n",
|
120 |
+
"hairclip: 23.44%\n",
|
121 |
+
"blunt bangs: 21.58%\n",
|
122 |
+
"picture frame: 19.34%\n",
|
123 |
+
"hand up: 18.26%\n",
|
124 |
+
"black skirt: 17.87%\n",
|
125 |
+
"smile: 17.87%\n",
|
126 |
+
"from behind: 13.57%\n",
|
127 |
+
"cowboy shot: 10.99%\n",
|
128 |
+
"indoors: 10.74%\n",
|
129 |
+
"curtains: 10.25%\n",
|
130 |
+
"facing away: 9.23%\n",
|
131 |
+
"white socks: 6.08%\n",
|
132 |
+
"bottle: 6.01%\n",
|
133 |
+
"mountain: 5.66%\n",
|
134 |
+
"blue skirt: 5.13%\n",
|
135 |
+
"drinking straw: 3.37%\n",
|
136 |
+
"kneehighs: 1.71%\n"
|
137 |
+
]
|
138 |
+
}
|
139 |
+
],
|
140 |
+
"source": [
|
141 |
+
"for tag, score in results:\n",
|
142 |
+
" print(f\"{tag}: {score*100:.2f}%\")"
|
143 |
+
]
|
144 |
+
}
|
145 |
+
],
|
146 |
+
"metadata": {
|
147 |
+
"kernelspec": {
|
148 |
+
"display_name": "py310",
|
149 |
+
"language": "python",
|
150 |
+
"name": "python3"
|
151 |
+
},
|
152 |
+
"language_info": {
|
153 |
+
"codemirror_mode": {
|
154 |
+
"name": "ipython",
|
155 |
+
"version": 3
|
156 |
+
},
|
157 |
+
"file_extension": ".py",
|
158 |
+
"mimetype": "text/x-python",
|
159 |
+
"name": "python",
|
160 |
+
"nbconvert_exporter": "python",
|
161 |
+
"pygments_lexer": "ipython3",
|
162 |
+
"version": "3.10.13"
|
163 |
+
}
|
164 |
+
},
|
165 |
+
"nbformat": 4,
|
166 |
+
"nbformat_minor": 2
|
167 |
+
}
|