Presentation is loading. Please wait.

Presentation is loading. Please wait.

Face Recognition with Deep Learning Method

Similar presentations


Presentation on theme: "Face Recognition with Deep Learning Method"— Presentation transcript:

1 Face Recognition with Deep Learning Method
Yucong Shen Face Recognition with Deep Learning Method

2 Labeled face in the wild (LFW)
LFW dataset contains images of faces collected in the web. Which has widely application in face verification. Mostly, LFW is a benchmark to test the performance of the algorithm in face recognition. In LFW dataset, there’re 5,749 identities, 1680 of the people pictured have two or more distinct photos in the data set. In the experiment, the input will be pair of images, the classifier will justify whether the two images are from the same person.

3 Labeled face in the wild
Sample Images from dataset In the paired images, mostly are the pictures from different views of same identity. All the images are in the wild so it’s approaching the real world application of face recognition.

4 键入说明。 Model: VGG-16 VGG-16: 16 weights layers, along with batch normalization, drop out. Two 3-by-3 filters has the same receptive field as one 5-by-5 filters, which makes VGG network has fewer parameters.

5 Training Input size: (150, 64, 64, 2).
Training Accuracy Input size: (150, 64, 64, 2). The optimization method: mini-batch stochastic gradient descent. Loss function: cross-entropy loss function with softmax function. learning rate: 0.01 Probability of drop out: 0.5. Training Loss

6 Experiment Result (Iter: 3000, C = .1) (Depth = 6, Num_tree = 200)
VGG-16 Test accuracy on LFW 55.06% 59.99% 85.59%


Download ppt "Face Recognition with Deep Learning Method"

Similar presentations


Ads by Google