Skip to content
Mijeong edited this page Oct 20, 2019 · 5 revisions

Dataset Mirror

[English]
MS COCO and Pascal VOC are the most popular datasets for studying for image classification, object detection, and segmentation. it takes a long time and effort to download a file because the data server is overseas. YOLK team created a data mirroring server for people studying in Korea. Mirroring Data Center is located in Seoul, South Korea. Download the datas a little faster here and save you time!
(You can check the download rate)

[Korean]
MS COCO와 Pascal VOCλŠ” 이미지 λΆ„λ₯˜, 객체 인식 그리고 μ„Έκ·Έλ©˜ν…Œμ΄μ…˜ 연ꡬ에 널리 μ‚¬μš©λ˜λŠ” λ°μ΄ν„°μ…‹μž…λ‹ˆλ‹€. ν•˜μ§€λ§Œ 데이터 μ„œλ²„κ°€ 해외에 μžˆμ–΄ νŒŒμΌμ„ λ‹€μš΄ λ°›λŠ”λ° 정말 였랜 μ‹œκ°„κ³Ό λ…Έλ ₯이 걸리죠 😩. YOLKνŒ€μ—μ„œλŠ” ν•œκ΅­μ—μ„œ λ”₯λŸ¬λ‹μ„ κ³΅λΆ€ν•˜λŠ” μ‚¬λžŒλ“€μ„ μœ„ν•΄ 데이터 미러링 μ„œλ²„λ₯Ό λ§Œλ“€μ—ˆμŠ΅λ‹ˆλ‹€! 미러링 데이터 μ„œλ²„λŠ” ν•œκ΅­, μ„œμšΈμ— μœ„μΉ˜ν•©λ‹ˆλ‹€. 쑰금 더 λΉ λ₯΄κ²Œ 데이터 νŒŒμΌμ„ λ‹€μš΄ λ°›μ•„ λ‹€μ–‘ν•œ λͺ¨λΈμ„ ν•™μŠ΅μ‹œμΌœλ³΄μ•„μš”! πŸ•Š
(λ‹€μš΄λ‘œλ“œ 속도 λΉ„κ΅λŠ” μ•„λž˜μͺ½μ—μ„œ 확인할 수 μžˆμŠ΅λ‹ˆλ‹€.)


MS COCO 2017

Train Data (19.3 GB)
Validation Date (0.77 GB)
Test Data (6.65 GB)
Annotation Data (0.25 GB)

from. Cocodataset.org

Pascal VOC

Train/Val 2012 Data(1.86 GB)
Test 2012 Data(1.72 GB)
Devkit2011

Train/Val 2007 Data(0.43 GB)
Test 2007 Data(1.36 GB)
Devkit2007

Benchmark(1.36 GB)

from.The PASCAL Visual Object Classes Homepage


Compare the Speed 😎

From Slow COCO DataCenter πŸ₯Ί

From Fast YOLK DataCenter X10πŸ₯°


Other cool datasets(ex. OpneImageV3, Kitty) will be added soon. Please look forward to! πŸ€—
If you have any questions, mail to [email protected] 🍳

Clone this wiki locally