Skip to content

Latest commit

 

History

History
25 lines (13 loc) · 771 Bytes

File metadata and controls

25 lines (13 loc) · 771 Bytes

Deep Learning For Species Recognition

This is a repository for sharing deep learning algorithms for recognizing species from images (including camera trap and drone images), such as Amur tiger, leopard, roe deer, et al.

Model

(1) Used transfer learning technique;

(2) Based on ResNet50 pre-trained from imagenet dataset;

(3) The last layer of ResNet50 is removed, and then from the second add the following layers:

(a) average pooling layer 2by2; 

(b) Use relu activation function to add a 1024 neuron fully connected layer; 

(c) Add an output layer with 3 neurons and softmax activation function.

Model Training

Model: optimizer: adam, loss function: 'categorical_crossentropy', metrics: accuracy.

Epochs: 30,

Batch_size: 32