Theme: Food Calorie Estimation
Team Members: Okan ALAN Gökberk Şahin Emre Yazıcı
This is our fourth blog post. In this blog post, we are examing methods.
Last week, we examined the dataset. We delete unnecessaries information, for example, in real features XML, there are names who created the dataset. We made the dataset usable.
METHODS for OBJECT DETECTION
During this week, we examined methods that are from related works. There are a lot of object detection methods. In this blog post, we will look Convolutional Neural Network(CNN).
Convolutional Neural Network
Convolutional Neural Networks are inspired by the brain. Research in the 1950s and 1960s by D.H Hubel and T.N Wiesel on the brain of mammals suggested a new model for how mammals perceive the world visually.
A CNN is very similar to ordinary Neural Networks. A CNN consists of an input and an output layer, as well as multiple hidden layers. The hidden layers of a CNN typically consist of convolutional layers, RELU layer i.e. activation function, pooling layers, fully connected layers, and normalization layers
Layers:[1]
- INPUT layer will hold the raw pixel values of the image.
- CONVOLUTION layer will compute the output of neurons that are connected to local regions in the input, each computing a dot product between their weights and a small region they are connected to in the input volume.
- RELU layer will apply an elementwise activation function, such as the max(0,x) thresholding at zero.
- POOLING(Subsampling in the above figure1) layer will perform a downsampling operation along the spatial dimensions
- FULLY-CONNECTED layer will compute the class scores, resulting in the volume of size.
Our Tried Methods
One of the object detection methods is Faster R-CNN. The researchers who inspired our project used this method. It is improved from CNN. We couldn’t run faster R-CNN. Therefore we studied on TensorFlow Detection Model.
We can detect fruits now but coin not. We researched and learned how to detect coin. Next week we are planning to detect coin. Then we will study for volume estimation.
