CV baseline之GoogLeNet v1





4个分支,1)有128个卷积核的1x1卷积  2)进行64个卷积核的1x1卷积,然后192个卷积核的3x3卷积 3)64个卷积核的1x1卷积,96个卷积核的5x5卷积  4)3x3 pool层,64个卷积核的1x1卷积

特征融合:GoogLeNet应用了赫布理论(the Hebbian principle:一起激发的神经元连在一起),Inception将各个分支所生成的特征图按照从左往右将通道加在一起,然后融合输出。其中,每个分支的特征图到紧挨在一起的,这也体现了赫布理论。



Despite concerns that max-pooling layers result in loss of accurate spatial information, the same convolutioanl network architecture has also been successfully employed for localization, object detection and human pose estimation.

The most straightforward way of improving the performance of deep neural networks is by increasing their size. Bigger size typically means a larger number of parameters, which makes the enlarged network more prone to overfitting.

3)为节省内存消耗,先将分辨率降低,再堆叠使用Inception module

For technical reasons(memory efficiency during training), it seemed beneficial to start using Inception modules only at higher layers while keeping the lower layers in traditional convolutional fashion.


We use an extra linear layer. This enables adapting and fine-tuning our networks for other label sets easily.


One interesting insight is that the strong performance of relatively shallower networks on this task suggests that the features produced by the layers in the middle of the network shoud be very discriminative.


fixed learning rate schedule (decreasing the learning rate by 4% every 8 epochs)


Still, one prescriotion that was verified to work very well after competition includes sampling of various sized pathces of the image whose size is distributied evenly between 8% and 100% of the image area and whose aspect ratio is chosed randomly between 3/4 and 4/3. Also, we found that the photometric distortions were useful to combat overfitting to some extent.


We started to use random interpolation methods(bilinear, area, nearest neighbor and cubic, with equal probability) for resizing relatively late and in conjuction with other hyperparameter changes.

9)实际应用中没必要用144 crops

We note that such agrresive cropping may not be necessary, in reall applications.



img: Golden Retriever from baidu.jpg is: golden retriever
207 n02099601 狗, golden retriever





posted @ 2020-08-05 15:38  sariel_sakura  阅读(230)  评论(0编辑  收藏  举报