Scribd
Upload
11 views

Transfer Learning Models.docx

Uploaded by

Manjiri Dhavtode
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
11 views

Transfer Learning Models.docx

Uploaded by

Manjiri Dhavtode
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

𝑇𝑟𝑎𝑛𝑠𝑓𝑒𝑟 𝐿𝑒𝑎𝑟𝑛𝑖𝑛𝑔 𝑀𝑜𝑑𝑒𝑙𝑠

● 𝐺𝑒𝑛𝑒𝑟𝑎𝑙𝑙𝑦 𝑤𝑒 𝑡𝑟𝑎𝑖𝑛𝑒𝑑 𝑡ℎ𝑒 𝑚𝑜𝑑𝑒𝑙 𝑜𝑛 𝑡ℎ𝑒 𝑔𝑖𝑣𝑒𝑛 𝑑𝑎𝑡𝑎


● 𝑀𝐿 𝑜𝑟 𝐷𝐿, 𝑤𝑒 𝑠𝑎𝑣𝑒 𝑡ℎ𝑒 𝑚𝑜𝑑𝑒𝑙 𝑖𝑛 𝑜𝑛𝑒 𝑓𝑜𝑟𝑚𝑎𝑡
● 𝑝𝑖𝑐𝑘𝑙𝑒 𝑜𝑟 𝐽𝑜𝑏 𝑙𝑖𝑏 𝐹𝑜𝑟 𝑀𝑎𝑐ℎ𝑖𝑛𝑒 𝑙𝑒𝑎𝑟𝑛𝑖𝑛𝑔
● 𝐷𝑒𝑒𝑝 𝑙𝑒𝑎𝑟𝑛𝑖𝑛𝑔 𝑡𝑓. 𝑠𝑎𝑣𝑒 𝑜𝑟 𝑢𝑠𝑖𝑛𝑔 𝑘𝑒𝑟𝑎𝑠 𝑎𝑙𝑠𝑜
● 𝐼𝑛 𝑡ℎ𝑒 𝑑𝑒𝑒𝑝 𝑙𝑒𝑎𝑟𝑛𝑖𝑛𝑔 𝑚𝑜𝑑𝑒𝑙 𝑠𝑎𝑣𝑖𝑛𝑔 𝑚𝑒𝑎𝑛𝑠 , 𝑠𝑎𝑣𝑖𝑛𝑔 𝑡ℎ𝑒 𝑤𝑒𝑖𝑔ℎ𝑡 𝑓𝑖𝑙𝑒
● 𝑌𝑜𝑢 𝑑𝑒𝑣𝑒𝑙𝑜𝑝 𝑡ℎ𝑒 𝑚𝑜𝑑𝑒𝑙 𝑤𝑖𝑡ℎ 𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙 𝑜𝑟𝑑𝑒𝑟
● 𝑇ℎ𝑖𝑠 𝑖𝑠 𝑐𝑎𝑙𝑙𝑒𝑑 𝑚𝑜𝑑𝑒𝑙 𝑐𝑜𝑛𝑓𝑖𝑔𝑢𝑟𝑎𝑡𝑖𝑜𝑛
● 𝑀𝑜𝑑𝑒𝑙 𝑐𝑜𝑛𝑓𝑖𝑔 ==== > 𝑀𝑜𝑑𝑒𝑙 𝑎𝑟𝑐ℎ𝑖𝑡𝑒𝑐𝑡𝑢𝑟𝑒
● 𝑀𝑜𝑑𝑒𝑙 𝑙𝑎𝑦𝑒𝑟𝑠 𝑚𝑒𝑎𝑛𝑠 ==== 𝑒𝑎𝑐ℎ 𝑙𝑎𝑦𝑒𝑟
● 𝑀𝑜𝑑𝑒𝑙 𝑠𝑢𝑚𝑚𝑎𝑟𝑦 ==== 𝑒𝑎𝑐ℎ 𝑙𝑎𝑦𝑒𝑟 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟𝑠
● 𝑀𝑜𝑑𝑒𝑙 𝑤𝑒𝑖𝑔ℎ𝑡𝑠 𝑚𝑒𝑎𝑛𝑠 ===== 𝑓𝑖𝑛𝑎𝑙 𝑤𝑒𝑖𝑔ℎ𝑡𝑠
● 𝑤𝑒 𝑑𝑒𝑣𝑒𝑙𝑜𝑝𝑒𝑑 𝑎𝑛𝑑 𝑤𝑒 𝑠𝑎𝑣𝑒𝑑: 𝐶𝑢𝑠𝑡𝑜𝑚 𝑚𝑜𝑑𝑒𝑙
● 𝐼 𝑔𝑖𝑣𝑒𝑛 𝑡ℎ𝑎𝑡 𝑠𝑎𝑣𝑒𝑑 𝑚𝑜𝑑𝑒𝑙 𝑡𝑜 𝑡ℎ𝑒 𝑐𝑙𝑖𝑛𝑒𝑡
● 𝐶𝑙𝑖𝑒𝑛𝑡 𝑤𝑖𝑙𝑙 𝑢𝑠𝑒 𝑡ℎ𝑎𝑡 𝑚𝑜𝑑𝑒𝑙 : 𝑇𝑟𝑎𝑛𝑠𝑓𝑒𝑟 𝑙𝑒𝑎𝑟𝑛𝑖𝑛𝑔
● 𝑇ℎ𝑒 𝑚𝑜𝑑𝑒𝑙𝑠 𝑤ℎ𝑖𝑐ℎ 𝑎𝑟𝑒 𝑢𝑠𝑒𝑖𝑛𝑔 𝑎𝑠 𝑇𝑟𝑎𝑛𝑠𝑓𝑒𝑟 𝑙𝑒𝑎𝑟𝑛𝑖𝑛𝑔 𝑐𝑎𝑙𝑙𝑒𝑑 𝑎𝑠 𝑃𝑟𝑒𝑡𝑟𝑎𝑖𝑛𝑒𝑑 𝑚𝑜𝑑𝑒𝑙𝑠
● 𝑃𝑟𝑒 𝑡𝑟𝑎𝑖𝑛𝑒𝑑 𝑚𝑒𝑎𝑛𝑠 𝑎𝑙𝑟𝑒𝑎𝑑𝑦 𝑡𝑟𝑎𝑖𝑛𝑒𝑑 𝑚𝑜𝑑𝑒𝑙

𝑁𝑜𝑤 𝑎 𝑑𝑎𝑦𝑠 𝑚𝑎𝑛𝑦 𝑐𝑜𝑚𝑝𝑎𝑛𝑖𝑒𝑠 𝑎𝑟𝑒 𝑡𝑟𝑎𝑖𝑛𝑖𝑛𝑔 𝑜𝑛 𝑒𝑛𝑒𝑚𝑜𝑟𝑜𝑢𝑠 𝑑𝑎𝑡𝑎

𝑎𝑛𝑑 𝑡ℎ𝑒𝑦 𝑎𝑟𝑒 𝑚𝑘𝑎𝑖𝑛𝑔 𝑝𝑢𝑏𝑙𝑖𝑐𝑙𝑦 𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑙𝑒 𝑚𝑜𝑑𝑒𝑙

𝑓𝑟𝑒𝑒 𝑜𝑟 𝑝𝑎𝑦 𝑎𝑠 𝑦𝑜𝑢 𝑔𝑜

𝐶𝑙𝑜𝑢𝑑 𝐴𝐼

𝐺𝑒𝑛𝐴𝐼

𝐶𝑁𝑁: 𝑂𝑏𝑗𝑒𝑐𝑡 𝑑𝑒𝑡𝑒𝑐𝑡𝑖𝑜𝑛

𝐿𝑒𝑁𝐸𝑇 ======== 𝑃𝑟𝑒𝑡𝑟𝑎𝑖𝑛𝑒𝑑 𝑚𝑜𝑑𝑒𝑙𝑠

𝐴𝑙𝑒𝑥𝑁𝑒𝑡

𝑉𝐺𝐺𝑛𝑒𝑡

𝑉𝐺𝐺16 − 32

𝑀𝑜𝑏𝑖𝑙𝑒𝑁𝑒𝑡

𝐺𝑜𝑜𝑔𝑙𝑒 𝑁𝑒𝑡

𝑅𝑒𝑠𝑁𝑒𝑡

𝐹𝑎𝑠𝑡𝑅𝐶𝑁𝑁

𝑀𝑎𝑠𝑘𝑅𝐶𝑁𝑁
𝐹𝑎𝑠𝑡𝑒𝑟𝑅𝐶𝑁𝑁

𝑌𝑂𝐿𝑂 𝑉8

● 𝐶𝑜𝑛𝑠𝑖𝑑𝑒𝑟 𝑀𝑜𝑏𝑖𝑙𝑒𝑁𝑒𝑡 𝑡𝑟𝑎𝑖𝑛𝑒𝑑 𝑜𝑛 1. 3𝑀𝑖𝑙𝑙𝑖𝑜𝑛 𝑖𝑚𝑎𝑔𝑒𝑠


● 𝐼𝑑𝑒𝑛𝑡𝑖𝑓𝑦 𝑡ℎ𝑒 𝑜𝑏𝑗𝑒𝑐𝑡𝑠 𝑎𝑟𝑜𝑢𝑛𝑑 1000 𝑐𝑙𝑎𝑠𝑠𝑒𝑠
● 𝑂𝑏𝑗𝑒𝑐𝑡 𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 𝑚𝑜𝑑𝑒𝑙
● 𝑤𝑒 ℎ𝑎𝑣𝑒 𝑡𝑤𝑜 𝑡𝑦𝑝𝑒𝑠 𝑜𝑓 𝑢𝑠𝑒 𝑐𝑎𝑠𝑒𝑠
o 𝑂𝑏𝑗𝑒𝑐𝑡 𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 : 𝑟𝑒𝑐𝑜𝑔𝑛𝑖𝑧𝑒 𝑡ℎ𝑒 𝑜𝑏𝑗𝑒𝑐𝑡 𝑎𝑛𝑑 𝑐𝑙𝑎𝑠𝑠𝑖𝑓𝑦 𝑎𝑙𝑠𝑜
o 𝑜𝑏𝑗𝑒𝑐ℎ 𝐷𝑒𝑡𝑒𝑐𝑡𝑖𝑜𝑛 : 𝐸𝑥𝑡𝑟𝑎𝑐𝑡 𝑡ℎ𝑒 𝑙𝑜𝑐𝑎𝑡𝑖𝑜𝑛 𝑎𝑙𝑠𝑜
● 𝑂𝑏𝑗𝑒𝑐𝑡 𝑑𝑒𝑡𝑒𝑐𝑡𝑖𝑜𝑛 𝑖𝑠 𝑢𝑠𝑒𝑑 𝑡𝑜 𝑒𝑥𝑡𝑟𝑎𝑐𝑡 𝑡ℎ𝑒 𝑐𝑜𝑜𝑟𝑑𝑖𝑛𝑎𝑡𝑒𝑠 𝑜𝑓 𝑜𝑏𝑗𝑒𝑐𝑡 𝑖𝑠 𝑐𝑎𝑙𝑙𝑒𝑑 𝑎𝑠 𝐵𝑜𝑢𝑛𝑑𝑖𝑛𝑔 𝑏𝑜𝑥
● 𝐵𝑜𝑢𝑛𝑑𝑖𝑛𝑔 𝑏𝑜𝑥 𝑚𝑒𝑎𝑛𝑠 𝑎 𝑟𝑒𝑐𝑡𝑎𝑛𝑔𝑢𝑙𝑎𝑟 𝑏𝑜𝑥 𝑜𝑛 𝑡ℎ𝑒 𝑖𝑚𝑎𝑔𝑒𝑠 , 𝑟𝑒𝑐𝑡𝑎𝑛𝑔𝑢𝑙𝑎𝑟 𝑏𝑜𝑥 𝑤𝑖𝑙𝑙 𝑔𝑖𝑣𝑒
(𝑥, 𝑦, 𝑥 + 𝑤, 𝑦 + ℎ)

𝑡𝑜𝑝 𝑙𝑒𝑓𝑡 𝑐𝑜𝑟𝑛𝑒𝑟 = (𝑥, 𝑦) 𝑏𝑜𝑡𝑡𝑜𝑚 𝑟𝑖𝑔ℎ𝑡 = (𝑥 + 𝑤, 𝑦 + ℎ)


𝐼𝑛𝑡𝑒𝑟 𝑠𝑒𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑈𝑛𝑖𝑜𝑛 (𝐼𝑂𝑈)

𝐴𝑐𝑡𝑢𝑎𝑙 𝑏𝑜𝑥𝑒𝑠 : 𝐺𝑟𝑜𝑢𝑛𝑑 𝑡𝑟𝑢𝑡ℎ

𝐺𝑟𝑜𝑢𝑛𝑑 𝑡𝑟𝑢𝑡ℎ 𝑏𝑏𝑜𝑥 𝑣𝑠 𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑜𝑛 𝑏𝑏𝑜𝑥


𝐹𝑖𝑛𝑒 𝑇𝑢𝑖𝑛𝑔

𝐶𝑜𝑛𝑠𝑖𝑑𝑒𝑟 𝑀𝑜𝑏𝑖𝑙𝑒𝑁𝑒𝑡 𝑡𝑟𝑎𝑖𝑛𝑒𝑑 𝑜𝑛 1. 3𝑀𝑖𝑙𝑙𝑖𝑜𝑛 𝑖𝑚𝑎𝑔𝑒𝑠

● 𝐼𝑑𝑒𝑛𝑡𝑖𝑓𝑦 𝑡ℎ𝑒 𝑜𝑏𝑗𝑒𝑐𝑡𝑠 𝑎𝑟𝑜𝑢𝑛𝑑 1000 𝑐𝑙𝑎𝑠𝑠𝑒𝑠


● 𝐼𝑚𝑎𝑔𝑖𝑛𝑒 𝑡ℎ𝑎𝑡 𝑦𝑜𝑢 𝑤𝑎𝑛𝑡 𝑖𝑑𝑒𝑛𝑖𝑡𝑓𝑦 𝑦𝑜𝑢𝑟 𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝐼𝐷 𝑐𝑎𝑟𝑑
● 𝑇ℎ𝑖𝑠 𝑚𝑜𝑑𝑒𝑙 𝑖𝑠 𝑛𝑜𝑡 𝑡𝑟𝑎𝑖𝑛𝑒𝑑 𝑜𝑛 𝑦𝑜𝑢𝑟 𝑐𝑙𝑔 𝑖𝑑 𝑐𝑎𝑟𝑑
● 𝑠𝑜 𝐼𝑡 𝑤𝑖𝑙𝑙 𝑛𝑜𝑡 𝑎𝑏𝑙𝑒 𝑡𝑜 𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑦 𝑡ℎ𝑒 𝑐𝑙𝑔 𝑖𝑑 𝑐𝑎𝑟𝑑
● 𝑌𝑜𝑢 𝑛𝑒𝑒𝑑 𝑡𝑜 𝑐𝑜𝑙𝑙𝑒𝑐𝑡 𝑠𝑜𝑚𝑒 𝑖𝑚𝑎𝑔𝑒𝑠 𝑜𝑓 𝑦𝑜𝑢𝑟 𝑐𝑙𝑔 𝑖𝑑 𝑐𝑎𝑟𝑑
● 𝑌𝑜𝑢 𝑛𝑒𝑒𝑑 𝑡𝑜 𝑐𝑟𝑒𝑎𝑡𝑒 𝑎 𝑑𝑎𝑡𝑎𝑠𝑒𝑡 : 𝑎𝑟𝑜𝑢𝑛𝑑 100 𝑖𝑚𝑎𝑔𝑒𝑠
● 𝐴𝑔𝑎𝑖𝑛 𝑦𝑜𝑢 𝑛𝑒𝑒𝑑 𝑡𝑜 𝑡𝑟𝑎𝑖𝑛 𝑡ℎ𝑒 𝑀𝑜𝑏𝑖𝑙𝑒𝑁𝑒𝑡 𝑚𝑜𝑑𝑒𝑙
● 𝑀𝑜𝑏𝑖𝑙𝑒 𝑛𝑒𝑡 𝑎𝑙𝑟𝑒𝑎𝑑𝑦 ℎ𝑎𝑠 1000 𝑐𝑙𝑎𝑠𝑠𝑒𝑠 + 1 𝑐𝑙𝑎𝑠𝑠𝑒𝑠
● 𝑎𝑐𝑐𝑜𝑟𝑑𝑖𝑛𝑔 𝑡𝑜 𝑡ℎ𝑖𝑠 𝑎𝑔𝑖𝑎𝑛 𝑤𝑒𝑖𝑔ℎ𝑡𝑠 𝑤𝑖𝑙 𝑏𝑒 𝑢𝑝𝑑𝑎𝑡𝑒𝑑
● 𝐹𝑖𝑛𝑒 𝑇𝑢𝑖𝑛𝑔

You might also like