1 × 1 convolutional with the softmax function  (SoftMax Inc)

Journal: IEEE access : practical innovations, open solutions

Article Title: Segmentation of Coronary Calcified Plaque in Intravascular OCT Images Using a Two-Step Deep Learning Approach

doi: 10.1109/access.2020.3045285

Figure Lengend Snippet: Illustration of SegNet architecture for calcium segmentation. The encoder is composed of a 3 × 3 convolution, batch normalization, and rectified linear unit layers. The decoder upsamples the low-resolution feature map using the transferred pooling indices from the counterpart encoder. The final output of decoder is fed to the Softmax activation to produce a pixel-wise classification map. The input is the preprocessed image selected by the classification model (step 1), and the output is predicted label. The sizes of input and output images are the same (200 × 448 pixels). In the input image, the black strip indicates the removed guidewire shadow.

Article Snippet: The restored feature map was fed to the final classification layer including a 1 × 1 convolution with Softmax activation to produce class probabilities for each pixel.

Techniques: Activation Assay, Stripping Membranes

Journal: GigaScience

Article Title: RootNav 2.0: Deep learning for automatic navigation of complex plant root architectures

doi: 10.1093/gigascience/giz123

Figure Lengend Snippet: The proposed CNN’s layers

Article Snippet: Softmax , Convolution (1 × 1) , 512 × 512 , 3.

Techniques: Blocking Assay