Places365-मानक डेटासेट में 365 दृश्य श्रेणियों से 1.8 मिलियन ट्रेन छवियां हैं, जिनका उपयोग Places365 CNNs को प्रशिक्षित करने के लिए किया जाता है। सत्यापन सेट में प्रति श्रेणी 50 चित्र और परीक्षण सेट में प्रति श्रेणी 900 चित्र हैं।
होमपेज : http://places2.csail.mit.edu/
स्रोत कोड :
tfds.datasets.places365_small.Builder
संस्करण :
-
2.1.0
(डिफ़ॉल्ट): केवाईडी के साथ एकीकरण को आसान बनाने के लिए उदाहरण कुंजियों को बदल दिया।
-
डाउनलोड आकार :
29.27 GiB
डेटासेट का आकार :
27.85 GiB
ऑटो-कैश्ड ( दस्तावेज़ीकरण ): नहीं
विभाजन :
विभाजित करना | उदाहरण |
---|---|
'test' | 328,500 |
'train' | 1,803,460 |
'validation' | 36,500 |
- फ़ीचर संरचना :
FeaturesDict({
'filename': Text(shape=(), dtype=string),
'image': Image(shape=(256, 256, 3), dtype=uint8),
'label': ClassLabel(shape=(), dtype=int64, num_classes=365),
})
- फ़ीचर दस्तावेज़ीकरण :
विशेषता | कक्षा | आकार | डीटाइप | विवरण |
---|---|---|---|---|
विशेषताएं डिक्ट | ||||
फ़ाइल का नाम | मूलपाठ | डोरी | ||
छवि | छवि | (256, 256, 3) | uint8 | |
लेबल | क्लासलेबल | int64 |
पर्यवेक्षित कुंजी (
as_supervised
दस्तावेज़ देखें):('image', 'label', 'filename')
चित्र ( tfds.show_examples ):
- उदाहरण ( tfds.as_dataframe ):
- उद्धरण :
@article{zhou2017places,
title={Places: A 10 million Image Database for Scene Recognition},
author={Zhou, Bolei and Lapedriza, Agata and Khosla, Aditya and Oliva, Aude and Torralba, Antonio},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
year={2017},
publisher={IEEE}
}