pyh5214 commited on
Commit
bdf42ae
ยท
verified ยท
1 Parent(s): 3f69cf5

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +38 -50
README.md CHANGED
@@ -74,59 +74,47 @@ predicted_class = torch.argmax(logits, dim=1)
74
 
75
 
76
  ## Training Details
77
- # Training Data:
78
 
79
- ๋“œ๋ก  ๋ฐ ์œ„์„ฑ ์ดฌ์˜ ๋‹ค์ค‘๋ถ„๊ด‘(5๋ฐด๋“œ) ์ด๋ฏธ์ง€ ๋ฐ์ดํ„ฐ์…‹
80
-
81
- ๋ผ๋ฒจ: ์ฃผ์š” ์ž‘๋ฌผ ๋ฐ ์‹์ƒ ํด๋ž˜์Šค
82
-
83
- # Training Procedure:
84
-
85
- ํŒŒ์ธํŠœ๋‹: facebook/convnext-tiny-224 ๊ธฐ๋ฐ˜
86
-
87
- ์—ํญ์ˆ˜: 30
88
-
89
- ๋ฐฐ์น˜์‚ฌ์ด์ฆˆ: 32
90
-
91
- ์˜ตํ‹ฐ๋งˆ์ด์ €: AdamW
92
-
93
- ํ•™์Šต๋ฅ : 1e-4, Cosine Annealing ์Šค์ผ€์ค„๋Ÿฌ ์‚ฌ์šฉ
94
-
95
- # Evaluation
96
- Testing Data: ๋ณ„๋„ ๋ณด์œ ํ•œ ๊ฒ€์ฆ์šฉ ๋‹ค์ค‘๋ถ„๊ด‘ ์ด๋ฏธ์ง€์…‹
97
-
98
- Metrics: ์ •ํ™•๋„(Accuracy), F1-Score
99
-
100
- Performance:
101
-
102
- Accuracy: 92.3%
103
-
104
- F1-Score: 0.91
105
-
106
- Environmental Impact
107
- Hardware: NVIDIA RTX 3090 GPU
108
-
109
- Training Duration: ์•ฝ 40์‹œ๊ฐ„
110
-
111
- Carbon Emissions: ์•ฝ 50 kg CO2e (ML CO2 ๊ณ„์‚ฐ๊ธฐ ๊ธฐ๋ฐ˜ ์ถ”์‚ฐ)
112
-
113
- Citation
114
- bibtex
115
- ๋ณต์‚ฌ
116
- @article{liu2022convnext,
117
- title={ConvNeXt: A ConvNet for the 2020s},
118
- author={Liu, Zhuang and Mao, Han and Wu, Chao and Feichtenhofer, Christoph and Darrell, Trevor and Xie, Saining},
119
- journal={arXiv preprint arXiv:2201.03545},
120
- year={2022}
121
  }
122
- Glossary
123
- ๋‹ค์ค‘๋ถ„๊ด‘ ์˜์ƒ(Multispectral Imagery): ์—ฌ๋Ÿฌ ํŒŒ์žฅ๋Œ€์˜ ๋น›์„ ๋ถ„๋ฆฌํ•˜์—ฌ ์ดฌ์˜ํ•œ ์˜์ƒ์œผ๋กœ, ์ž‘๋ฌผ์˜ ์ƒ์œก ์ƒํƒœ ๋ถ„์„ ๋“ฑ์— ํ™œ์šฉ๋จ
124
 
125
- ConvNeXt: ํ˜„๋Œ€์ ์ธ ๊ตฌ์กฐ๋ฅผ ๊ฐ–์ถ˜ ์ปจ๋ณผ๋ฃจ์…˜ ์‹ ๊ฒฝ๋ง(CNN)
126
 
127
- F1-Score: ์ •๋ฐ€๋„์™€ ์žฌํ˜„์œจ์˜ ์กฐํ™”ํ‰๊ท , ๋ถˆ๊ท ํ˜• ๋ฐ์ดํ„ฐ ํ‰๊ฐ€์— ํšจ๊ณผ์ 
 
 
128
 
129
- Model Card Authors
130
- AI Research Team, Your Organization
131
 
132
 
 
74
 
75
 
76
  ## Training Details
 
77
 
78
+ - **Training Data:**
79
+ - ๋“œ๋ก  ๋ฐ ์œ„์„ฑ ์ดฌ์˜ ๋‹ค์ค‘๋ถ„๊ด‘(5๋ฐด๋“œ) ์ด๋ฏธ์ง€ ๋ฐ์ดํ„ฐ์…‹
80
+ - ๋ผ๋ฒจ: ์ฃผ์š” ์ž‘๋ฌผ ๋ฐ ์‹์ƒ ํด๋ž˜์Šค
81
+ - **Training Procedure:**
82
+ - ํŒŒ์ธํŠœ๋‹: facebook/convnext-tiny-224 ๊ธฐ๋ฐ˜
83
+ - ์—ํญ์ˆ˜: 30
84
+ - ๋ฐฐ์น˜์‚ฌ์ด์ฆˆ: 32
85
+ - ์˜ตํ‹ฐ๋งˆ์ด์ €: AdamW
86
+ - ํ•™์Šต๋ฅ : 1e-4, Cosine Annealing ์Šค์ผ€์ค„๋Ÿฌ ์‚ฌ์šฉ
87
+
88
+ ## Evaluation
89
+
90
+ - **Testing Data:** ๋ณ„๋„ ๋ณด์œ ํ•œ ๊ฒ€์ฆ์šฉ ๋‹ค์ค‘๋ถ„๊ด‘ ์ด๋ฏธ์ง€์…‹
91
+ - **Metrics:** ์ •ํ™•๋„(Accuracy), F1-Score
92
+ - **Performance:**
93
+ - Accuracy: 92.3%
94
+ - F1-Score: 0.91
95
+
96
+ ## Environmental Impact
97
+
98
+ - **Hardware:** NVIDIA RTX 3090 GPU
99
+ - **Training Duration:** ์•ฝ 40์‹œ๊ฐ„
100
+ - **Carbon Emissions:** ์•ฝ 50 kg CO2e (ML CO2 ๊ณ„์‚ฐ๊ธฐ ๊ธฐ๋ฐ˜ ์ถ”์‚ฐ)
101
+
102
+ ## Citation
103
+
104
+ @article{liu2022convnext,
105
+ title={ConvNeXt: A ConvNet for the 2020s},
106
+ author={Liu, Zhuang and Mao, Han and Wu, Chao and Feichtenhofer, Christoph and Darrell, Trevor and Xie, Saining},
107
+ journal={arXiv preprint arXiv:2201.03545},
108
+ year={2022}
 
 
 
 
 
 
 
 
 
 
 
109
  }
 
 
110
 
111
+ ## Glossary
112
 
113
+ - **๋‹ค์ค‘๋ถ„๊ด‘ ์˜์ƒ(Multispectral Imagery):** ์—ฌ๋Ÿฌ ํŒŒ์žฅ๋Œ€์˜ ๋น›์„ ๋ถ„๋ฆฌํ•˜์—ฌ ์ดฌ์˜ํ•œ ์˜์ƒ์œผ๋กœ, ์ž‘๋ฌผ์˜ ์ƒ์œก ์ƒํƒœ ๋ถ„์„ ๋“ฑ์— ํ™œ์šฉ๋จ
114
+ - **ConvNeXt:** ํ˜„๋Œ€์ ์ธ ๊ตฌ์กฐ๋ฅผ ๊ฐ–์ถ˜ ์ปจ๋ณผ๋ฃจ์…˜ ์‹ ๊ฒฝ๋ง(CNN)
115
+ - **F1-Score:** ์ •๋ฐ€๋„์™€ ์žฌํ˜„์œจ์˜ ์กฐํ™”ํ‰๊ท , ๋ถˆ๊ท ํ˜• ๋ฐ์ดํ„ฐ ํ‰๊ฐ€์— ํšจ๊ณผ์ 
116
 
117
+ ## Model Card Authors
 
118
 
119
+ - AI Research Team, Your Organization
120