Global Leaderboard Ranking (Indian Pines)
์๋ก์ด๋ฉ์ Hyperspectral Image Classification์ ๊ธฐ์กด ๊ธ๋ก๋ฒ(SOTA) ๊ธฐ๋ก์ ๊ฒฝ์ ํ๋ฉฐ 1์์ ๋ฑ์ฌ๋์์ต๋๋ค. ELROILABโs Hyperspectral Image Classification has achieved the top rank (1st place) by surpassing existing global State-of-the-Art (SOTA) records.
| Rank | Model Name | Accuracy | Organization | Status |
|---|---|---|---|---|
| ๐ฅ 1st | SC-DBNET(ELROILAB) | 97.84% | ELROILAB | Current SOTA |
| ๐ฅ 2nd | PRNET-RF | 90.23% | Academic (Ref.) | - |
| ๐ฅ 3rd | HYLITE | 89.80% | Academic (Ref.) | - |
ELROILAB: SOTA #1 Ranking in Hyperspectral Image Classification
[SOTA #1 Ranking] Indian Pines Hyperspectral Classification
ELROILAB: Hyperspectral Image Classification Ranked 1st(SOTA) on Indian Pines Benchmark
License Notice: ๋ณธ ๋ชจ๋ธ์ ์์ ๊ถ์ (์ฃผ)์๋ก์ด๋ฉ์ ์์ผ๋ฉฐ, ์์ ์ ์ด์ฉ ๋ฐ ๋ฌด๋จ ๋ฐฐํฌ๋ฅผ ๊ธํฉ๋๋ค. (The ownership of this model belongs to ELROILAB Co., Ltd. Commercial use and unauthorized distribution are strictly prohibited.) (Proprietary License)
[KOREAN]
์๋ก์ด๋ฉ(ELROILAB)์ ๋น์ ํ ์ด๋ฌผ์ ๊ฒ์ถํ๋ AI ์ด๋ถ๊ด(AI Hyperspectral) ์์ฒ ๊ธฐ์ ์ ๋ณด์ ํ๊ณ ์์ต๋๋ค. ๋ณธ ๋ชจ๋ธ ์นด๋๋ ์๋ก์ด๋ฉ ์์ฒด ๊ฒ์ฆ ๋ฐ์ดํฐ๋ฅผ ๊ธฐ๋ฐ์ผ๋ก ํ ์ฑ๋ฅ ์งํ๋ฅผ ๋ด๊ณ ์์ต๋๋ค.
ํต์ฌ ์ฑ๋ฅ (Performance)
์๋ก์ด๋ฉ, ์์ฌ ์ด๋ถ๊ด ์ด๋ฏธ์ง ๋ถ๋ฅ๋ชจ๋ธ "SC-DBNET: AN EFFICIENT SPECTRAL-CONDITIONED DUAL-BRANCH NET FOR HSI CLASSIFICATION" ์ธ๊ณ ์ต๊ณ ๊ถ์์ ์ง๊ตฌ๊ณผํ ๋ฐ ์๊ฒฉ ํ์ฌ ํ์ ๋ํ IGARSS 2026์ ์ต์ข ์ฑํ ๋์์ต๋๋ค. ๊ธ๋ก๋ฒ AI ํ๋ธ์ธ Hugging Face์ ๊ณต์ ๊ท๊ฒฉ์ ๋ฐ๋ผ ๋ฑ์ฌ๋ ์๋ก์ด๋ฉ์ ์์ฒด ๊ฒ์ฆ ๊ฒฐ๊ณผ์ ๋๋ค.
- ์ฃผ์ ์์น: Hyperspectral Image Classification 97.84%(Overall Accuracy ๊ธฐ์ค), SOTA Ranking 1์ ๋ฌ์ฑ
- ๊ฒ์ฆ ๊ธฐ์ค: Indian Pines (์ธ๋์ธ ํ์ธ์ฆ) ๋ฐ์ดํฐ์ ๊ธฐ์ค ํ์ค์ํ ๋ฒค์น๋งํฌ
๊ธฐ์ ์ ์ฐจ๋ณ์ (SOTA Evidence)
์๋ก์ด๋ฉ์ AI๋ "๋ฌผ์ง ๊ณ ์ ์ ๋ถ๊ด ํน์ฑ(Spectral Signature)์ ์ฝ๋ ์ด๋ถ๊ด(Hyperspectral)"์ ๋ถ์ํ์ฌ 97.84%(์ธ๊ณ์ต๊ณ ์์ค) ์ด๋ถ๊ด ์ด๋ฏธ์ง ๋ถ๋ฅ ์ฑ๋ฅ์ ๋ณด์ฌ์ค๋๋ค.
[ENGLISH]
ELROILAB holds proprietary AI Hyperspectral technology capable of detecting irregular foreign objects. This Model Card contains performance metrics based on ELROILABโs internal validation data.
Key Performance
ELROILABโs proprietary hyperspectral image classification model, "SC-DBNet: An Efficient Spectral-Conditioned Dual-Branch Net for HSI Classification," has been officially accepted for IGARSS 2026, the worldโs most prestigious conference in geoscience and remote sensing. The following performance indicators have been officially registered on Hugging Face, the global AI hub, in accordance with its standard specifications.
- Primary Metric: 97.84% (Overall Accuracy) in Hyperspectral Image Classification, Achieved 1st Place in SOTA Rankings
- Benchmark: Standard test performed on the Indian Pines dataset
Technical Differentiation (SOTA Evidence)
ELROILABโs AI analyzes "Spectral Signatures," the unique material properties captured through hyperspectral imaging, delivering world-class classification performance with an accuracy of 97.84%.
[๋น์ฆ๋์ค ๋ฌธ์ ๋ฐ ์์ธ ๋ฆฌํฌํธ ์์ฒญ, INQUIRY & REPORT REQUEST]
- ํํ์ด์ง(Web): www.elroilab.com
- ์ด๋ฉ์ผ(E-Mail): marketing@elroilab.com
Evaluation results
- Overall Accuracy (SOTA) on Indian Pinesself-reported97.840
