- cuda resnet18_cbam 只有訓練最後一層fc conv層凍結 - batch32 epoch22  - cuda resnet18_cbam 有pre train 但conv層沒凍結 - batch32 epoch22  - cuda resnet50_cbam 有pre train - conv層沒凍結batch 32 epoch 22   - cuda resnet50_cbam 有pre train - 但conv層沒凍結batch 32 epoch 30   - cuda resnet50_cbam 有pre train - 但conv層沒凍結batch 128 epoch 20 有影像增強(隨機水平翻轉 隨機色調變化 隨機旋轉)   - cuda resnet152_cbam 有pre train - 但conv層沒凍結batch 128 epoch 20 有影像增強(隨機水平翻轉 隨機色調變化 隨機旋轉)   - cuda resnet152_cbam 有pre train - 凍結layer1 batch 32 epoch 20 有影像增強(隨機水平翻轉 隨機色調變化 隨機旋轉)   - cuda resnet152_cbam 有pre train - 凍結layer1 batch 64 epoch 20 有影像增強(隨機水平翻轉 隨機色調變化 隨機旋轉)   - cuda resnet50_cbam 有pre train - 凍結layer1 batch 64 epoch 20 有影像增強(隨機水平翻轉 隨機色調變化 隨機旋轉)   - cuda resnet152_cbam 有pre train - 凍結layer1 batch 64 epoch 20 有影像增強(隨機水平翻轉 隨機色調變化 隨機旋轉)   - cuda resnet152_cbam 有pre train - 經過高斯模糊+灰階處理 batch 32 epoch 20 有影像增強(隨機水平翻轉 隨機旋轉)   - cuda resnet152_cbam 有pre train - 經過高斯模糊+灰階處理 batch 32 epoch 50 有影像增強(隨機水平翻轉 隨機旋轉)   - cuda resnet152_cbam 有pre train (沒凍結層) - 經過高斯模糊+灰階處理 batch 32 epoch 50 有影像增強!(隨機水平翻轉 隨機旋轉)   ### 資料前處理 --- 高斯模糊+灰階 https://www.kaggle.com/code/samkit5/final-project-dr-detection-512 其他都是pytorch 內建 ### 老錢代紀錄 --- #### resnet152+cbam - Epoch = 25 - 擴充坐滿  #### resnet50+cbam - Epoch = 15 - 擴充坐滿   #### effeicientnet  
×
Sign in
Email
Password
Forgot password
or
By clicking below, you agree to our
terms of service
.
Sign in via Facebook
Sign in via Twitter
Sign in via GitHub
Sign in via Dropbox
Sign in with Wallet
Wallet (
)
Connect another wallet
New to HackMD?
Sign up