ragavsachdeva commited on
Commit
bd1dc48
1 Parent(s): f7499c0

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +141 -0
README.md ADDED
@@ -0,0 +1,141 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ tags:
5
+ - Manga
6
+ - Object Detection
7
+ - OCR
8
+ - Clustering
9
+ - Diarisation
10
+ ---
11
+ <style>
12
+ .title-container {
13
+ display: flex;
14
+ flex-direction: column; /* Stack elements vertically */
15
+ justify-content: center;
16
+ align-items: center;
17
+ }
18
+
19
+ .title {
20
+ font-size: 2em;
21
+ text-align: center;
22
+ color: #333;
23
+ font-family: 'Comic Sans MS', cursive; /* Use Comic Sans MS font */
24
+ text-transform: uppercase;
25
+ letter-spacing: 0.1em;
26
+ padding: 0.5em 0 0.2em;
27
+ background: transparent;
28
+ }
29
+
30
+ .title span {
31
+ background: -webkit-linear-gradient(45deg, #6495ED, #4169E1); /* Blue gradient */
32
+ -webkit-background-clip: text;
33
+ -webkit-text-fill-color: transparent;
34
+ }
35
+
36
+ .subheading {
37
+ font-size: 1.5em; /* Adjust the size as needed */
38
+ text-align: center;
39
+ color: #555; /* Adjust the color as needed */
40
+ font-family: 'Comic Sans MS', cursive; /* Use Comic Sans MS font */
41
+ }
42
+
43
+ .authors {
44
+ font-size: 1em; /* Adjust the size as needed */
45
+ text-align: center;
46
+ color: #777; /* Adjust the color as needed */
47
+ font-family: 'Comic Sans MS', cursive; /* Use Comic Sans MS font */
48
+ padding-top: 1em;
49
+ }
50
+
51
+ .affil {
52
+ font-size: 1em; /* Adjust the size as needed */
53
+ text-align: center;
54
+ color: #777; /* Adjust the color as needed */
55
+ font-family: 'Comic Sans MS', cursive; /* Use Comic Sans MS font */
56
+ }
57
+
58
+ </style>
59
+
60
+ <div class="title-container">
61
+ <div class="title">
62
+ Ta<span>il</span>s Tell Ta<span>le</span>s
63
+ </div>
64
+ <div class="subheading">
65
+ Chapter-Wide Manga Transcriptions With Character Names
66
+ </div>
67
+ <div class="authors">
68
+ Ragav Sachdeva, Gyungin Shin and Andrew Zisserman
69
+ </div>
70
+ <div class="affil">
71
+ University of Oxford
72
+ </div>
73
+ <div style="display: flex;">
74
+ <a href="https://arxiv.org/abs/2408.00298"><img alt="Static Badge" src="https://img.shields.io/badge/arXiv-2408.00298-blue"></a>
75
+ &emsp;
76
+ <img alt="Dynamic JSON Badge" src="https://img.shields.io/badge/dynamic/json?url=https%3A%2F%2Fhuggingface.co%2Fapi%2Fmodels%2Fragavsachdeva%2Fmagiv2%3Fexpand%255B%255D%3Ddownloads%26expand%255B%255D%3DdownloadsAllTime&query=%24.downloadsAllTime&label=%F0%9F%A4%97%20Downloads">
77
+ </div>
78
+ </div>
79
+
80
+
81
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/630852d2f0dc38fb47c347a4/OQW4r_A3aA9RrWpG6Wkve.png)
82
+
83
+ # Usage
84
+ ```python
85
+ from PIL import Image
86
+ import numpy as np
87
+ from transformers import AutoModel
88
+ import torch
89
+
90
+ model = AutoModel.from_pretrained("ragavsachdeva/magiv2", trust_remote_code=True).cuda().eval()
91
+
92
+
93
+ def read_image(path_to_image):
94
+ with open(path_to_image, "rb") as file:
95
+ image = Image.open(file).convert("L").convert("RGB")
96
+ image = np.array(image)
97
+ return image
98
+
99
+ chapter_pages = ["page1.png", "page2.png", "page3.png" ...]
100
+ character_bank = {
101
+ "images": ["char1.png", "char2.png", "char3.png", "char4.png" ...],
102
+ "names": ["Luffy", "Sanji", "Zoro", "Ussop" ...]
103
+ }
104
+
105
+ chapter_pages = [read_image(x) for x in chapter_pages]
106
+ character_bank["images"] = [read_image(x) for x in character_bank["images"]]
107
+
108
+ with torch.no_grad():
109
+ per_page_results = model.do_chapter_wide_prediction(chapter_pages, character_bank, use_tqdm=True, do_ocr=True)
110
+
111
+ transcript = []
112
+ for i, (image, page_result) in enumerate(zip(chapter_pages, per_page_results)):
113
+ model.visualise_single_image_prediction(image, page_result, f"page_{i}.png")
114
+ speaker_name = {
115
+ text_idx: page_result["character_names"][char_idx] for text_idx, char_idx in page_result["text_character_associations"]
116
+ }
117
+ for j in range(len(page_result["ocr"])):
118
+ if not page_result["is_essential_text"][j]:
119
+ continue
120
+ name = speaker_name.get(j, "unsure")
121
+ transcript.append(f"<{name}>: {page_result['ocr'][j]}")
122
+ with open(f"transcript.txt", "w") as fh:
123
+ for line in transcript:
124
+ fh.write(line + "\n")
125
+ ```
126
+
127
+ # License and Citation
128
+ The provided model and datasets are available for unrestricted use in personal, research, non-commercial, and not-for-profit endeavors. For any other usage scenarios, kindly contact me via email, providing a detailed description of your requirements, to establish a tailored licensing arrangement.
129
+ My contact information can be found on my website: ragavsachdeva [dot] github [dot] io
130
+
131
+ ```
132
+ @misc{magiv2,
133
+ title={Tails Tell Tales: Chapter-Wide Manga Transcriptions with Character Names},
134
+ author={Ragav Sachdeva and Gyungin Shin and Andrew Zisserman},
135
+ year={2024},
136
+ eprint={2408.00298},
137
+ archivePrefix={arXiv},
138
+ primaryClass={cs.CV},
139
+ url={https://arxiv.org/abs/2408.00298},
140
+ }
141
+ ```