How Can I combine HOG descriptors as input and CNN for Classification as output

9 次查看（过去 30 天）
Matpar
Matpar 2020-7-12
评论： Shams Shams 2023-1-4
Hi all professionals and good day,
in reference to my question, is it possible?
I have no code just yet! my process at the moment is to understand if this can be done and what are the procedures involved so that I can follow such step by step. I tried googling but I am sure that I have the wrong terminology and idea of the procedure!
Conundrum:
I am trying to extract HOG descriptor features from video sequences and feed such into rcnn for classification as the output!
Question:
  1. is that possible?
  2. if yes, can you point me in the right direction, please? (links , example of processes etc)
I would like to learn this on my own so that i can get the gist of the procedures involved!
what I don't know is the procedure to start the code generation process!
Please assist!
thank you in advance for your assistance and for responding to my absurd questions!

请先登录，再回答此问题。

回答（4 个）

Fadi Alsuhimat
Fadi Alsuhimat 2023-1-3
You can combine the features with dense layers. Example:
  • CNN model has global average pooling before full connected layers. (512).
  • Extracted hog feature vector size is 512 . You pass this feature into (512) dense layers. concatenate cnn fully connected layer with hog feature layer.
  • After concatenate include one more fully connected layer and drop out. Include one more dense layer with softmax or sigmoid .
Mahesh Taparia
Mahesh Taparia 2020-7-17
Hi
It seems you want to classify the image based on HOG features. You can do by storing the features in an array and pass it to a CNN. You can refer to this documentation which explain the classification of image using CNN. Moreover, you can refer to this documentation for other approach which is taking HOG features and classify the images using SVM. Hope it will helps!
  2 个评论
Kinjal Joshi
Kinjal Joshi 2021-2-10
How HOG features can be passed to CNN?
I want to use both features i.e. extracted by CNN and HOG features with use of concatenate layer. featureinput layer takes only dimension of features. How actual HOG features can be given to CNN?

请先登录，再进行评论。

Fadi Alsuhimat
Fadi Alsuhimat 2023-1-3
You can combine the features with dense layers. Example:
  • CNN model has global average pooling before full connected layers. (512).
  • Extracted hog feature vector size is 512 . You pass this feature into (512) dense layers. concatenate cnn fully connected layer with hog feature layer.
  • After concatenate include one more fully connected layer and drop out. Include one more dense layer with softmax or sigmoid .
Fadi Alsuhimat
Fadi Alsuhimat 2023-1-3
You can combine the features with dense layers. Example:
  • CNN model has global average pooling before full connected layers. (512).
  • Extracted hog feature vector size is 512 . You pass this feature into (512) dense layers. concatenate cnn fully connected layer with hog feature layer.
  • After concatenate include one more fully connected layer and drop out. Include one more dense layer with softmax or sigmoid .

类别

在 Help Center 和 File Exchange 中查找有关 Object Detection 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by